
%%html
<style>
    .qst {padding:10px; border-radius: 5px; border: solid 2px #EF6E1A;}
    .qst:before {font-weight: bold;display: block; margin: 0px 10px 10px 10px;}
    h1,h3 {color: #EF6E1A;}
</style>


#!pip install imbalanced-learn
#!pip install mlxtend

# -- Import necessary libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
import plotly.graph_objects as go
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.offline as pyo
import json
import sys

pyo.init_notebook_mode(connected=True)
#pio.renderers.default = "notebook"
pd.set_option('display.max_columns', None)

# ML functions
from ML_functions.plots import plot_histogram
from ML_functions.util_functions import plot_diagnostic_outlier,select_diagnostic_outliers
from ML_functions.plots import plot_pie, plot_time_series, matrix_correlation

# --- Model functions

from imblearn.pipeline import Pipeline as ImbPipeline
from imblearn.over_sampling import SMOTE
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import RobustScaler


from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
import category_encoders as ce
import warnings
from sklearn.pipeline import make_pipeline
from sklearn.model_selection import GridSearchCV

# Performance Metrics

from sklearn.metrics import accuracy_score,roc_auc_score
from sklearn.metrics import confusion_matrix,classification_report
from mlxtend.plotting import plot_confusion_matrix
from sklearn.metrics import precision_score, recall_score
from sklearn.metrics import f1_score
import matplotlib.pyplot as plt

# ======== LIBRERIAS MODELLING =======
from sklearn.linear_model import LogisticRegression
import  xgboost as xgb
from sklearn.ensemble import BaggingClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier
from sklearn.dummy import DummyClassifier

warnings.filterwarnings("ignore")


# --- Load the datasets
customer_loan_df = pd.read_parquet("input/main_dataset.parquet") #---> main dataset
credit_report_df = pd.read_parquet("input/credit_reports.parquet") # ---> credit reports


# --- Examine the first columns of the customer dataset
print(f"The shape of the dataset is: {customer_loan_df.shape}")
customer_loan_df.head(7)

The shape of the dataset is: (14454, 17)


# --- Examine the first columns of the customer dataset
print(f"The shape of the dataset is: {credit_report_df.shape}")
credit_report_df.head(7)

The shape of the dataset is: (287356, 29)


# --- Examine the basic statistics of the customer_loan_df
customer_loan_df.describe().T


# --- Examine the basic statistics of the customer_loan_df
credit_report_df.describe().T


# --- Merge the datasets using the customer_id column
bankaya_merged = pd.merge(customer_loan_df,credit_report_df, on='customer_id',how='inner')

# --- Examine the first columns of the new dataset
print(f"The shape of the merged dataset is: {bankaya_merged.shape}")
print(f'previous customers amount: {len(customer_loan_df.customer_id.unique())}')
print(f'current customers amount: {len(bankaya_merged.customer_id.unique())}')
bankaya_merged.head(5)

The shape of the merged dataset is: (287356, 45)
previous customers amount: 14454
current customers amount: 9249


# --- Examine the basic statistics of the merged dataframe
bankaya_merged.describe().T


from ML_functions.util_functions import inspection


# --- Inspect the Dataframe in order to determine cardinality, missing values, encoding strategy, imputing and transforming variables. 
inspection(bankaya_merged)

==== Number of labeles: Cardinality ====

ACCOUNT_TYPE contains 7 labels
ACC_CREATION_DATETIME contains 9236 labels
APPLICATION_DATETIME contains 368 labels
BUSINESS_TYPE contains 59 labels
CDC_INQUIRY_ID contains 9173 labels
CREDIT_TYPE contains 34 labels
CURRENCY contains 4 labels
CURRENT_PAYMENT contains 87 labels
DELAYED_PAYMENTS contains 98 labels
INQUIRY_DATE contains 626 labels
INQUIRY_TIME contains 9261 labels
LAST_PAYMENT_DATE contains 7269 labels
LAST_PURCHASE_DATE contains 7711 labels
LOAN_CLOSING_DATE contains 7168 labels
LOAN_OPENING_DATE contains 7944 labels
LOAN_ORIGINATION_DATETIME contains 9005 labels
PAYMENT_FREQUENCY contains 12 labels
PREVENTION_KEY contains 32 labels
REPORT_DATE contains 5814 labels
RESPONSABILITY_TYPE contains 6 labels
TOTAL_REPORTED_PAYMENTS contains 2 labels
UPDATE_DATE contains 5822 labels
WORST_DELAY_DATE contains 5692 labels
first_app_date contains 682 labels
first_bnpl_app_date contains 8507 labels
last_app_date contains 684 labels
last_bnpl_app_date contains 8507 labels
 
==== Types of the variables we are working with ====

customer_id                                  int64
loan_id                                      int64
ACC_CREATION_DATETIME               datetime64[ns]
APPLICATION_DATETIME                datetime64[ns]
LOAN_ORIGINATION_DATETIME           datetime64[ns]
max_days_late                                int64
target                                       int64
account_to_application_days                  int64
n_sf_apps                                  float64
first_app_date                      datetime64[ns]
last_app_date                       datetime64[ns]
n_bnpl_apps                                float64
n_bnpl_approved_apps                       float64
first_bnpl_app_date                 datetime64[ns]
last_bnpl_app_date                  datetime64[ns]
n_inquiries_l3m                            float64
n_inquiries_l6m                            float64
INQUIRY_TIME                   datetime64[ns, UTC]
CDC_INQUIRY_ID                              object
INQUIRY_DATE                        datetime64[ns]
PREVENTION_KEY                              object
CURRENCY                                    object
MAX_CREDIT                                 float64
CREDIT_LIMIT                               float64
PAYMENT_AMOUNT                             float64
UPDATE_DATE                         datetime64[ns]
LOAN_OPENING_DATE                   datetime64[ns]
LOAN_CLOSING_DATE                   datetime64[ns]
WORST_DELAY_DATE                    datetime64[ns]
REPORT_DATE                         datetime64[ns]
LAST_PURCHASE_DATE                  datetime64[ns]
LAST_PAYMENT_DATE                   datetime64[ns]
PAYMENT_FREQUENCY                           object
BUSINESS_TYPE                               object
CREDIT_TYPE                                 object
ACCOUNT_TYPE                                object
RESPONSABILITY_TYPE                         object
TOTAL_PAYMENTS                             float64
DELAYED_PAYMENTS                            object
CURRENT_PAYMENT                             object
WORST_DELAY                                float64
TOTAL_REPORTED_PAYMENTS                     object
CURRENT_BALANCE                            float64
BALANCE_DUE                                float64
BALANCE_DUE_WORST_DELAY                    float64
dtype: object 


Total Samples with missing values: 267458

====Total percentage of missing values per variable ====

WORST_DELAY_DATE               70.538983
LOAN_CLOSING_DATE              32.971993
last_app_date                  29.469369
n_sf_apps                      29.469369
first_app_date                 29.469369
TOTAL_REPORTED_PAYMENTS        14.595484
LAST_PAYMENT_DATE              11.023608
first_bnpl_app_date             7.033784
n_bnpl_approved_apps            7.033784
n_bnpl_apps                     7.033784
last_bnpl_app_date              7.033784
TOTAL_PAYMENTS                  6.488467
CREDIT_LIMIT                    2.908239
LAST_PURCHASE_DATE              1.441418
WORST_DELAY                     1.117081
n_inquiries_l3m                 0.591601
n_inquiries_l6m                 0.591601
MAX_CREDIT                      0.063336
PAYMENT_FREQUENCY               0.030972
REPORT_DATE                     0.030972
BALANCE_DUE                     0.030972
DELAYED_PAYMENTS                0.030972
BUSINESS_TYPE                   0.030972
CREDIT_TYPE                     0.030972
CURRENT_PAYMENT                 0.030972
RESPONSABILITY_TYPE             0.030972
CURRENT_BALANCE                 0.030972
ACCOUNT_TYPE                    0.030972
BALANCE_DUE_WORST_DELAY         0.030972
LOAN_OPENING_DATE               0.030972
UPDATE_DATE                     0.030972
PAYMENT_AMOUNT                  0.030972
CURRENCY                        0.030972
PREVENTION_KEY                  0.030972
INQUIRY_DATE                    0.030972
CDC_INQUIRY_ID                  0.030972
loan_id                         0.000000
INQUIRY_TIME                    0.000000
account_to_application_days     0.000000
target                          0.000000
max_days_late                   0.000000
LOAN_ORIGINATION_DATETIME       0.000000
APPLICATION_DATETIME            0.000000
ACC_CREATION_DATETIME           0.000000
customer_id                     0.000000
dtype: float64

 ===== Removed 2119 duplicate rows.======


# --- Replace None with the median value of the column
bankaya_merged['DELAYED_PAYMENTS'].fillna(bankaya_merged['DELAYED_PAYMENTS'].median(), inplace=True)
# --- convert the column to a int. 
bankaya_merged['DELAYED_PAYMENTS'] = bankaya_merged['DELAYED_PAYMENTS'].astype('int')


# --- Duplicate the Ddatframe to do the EDA
bankaya_eda = bankaya_merged.copy()


#bankaya_merged.head(5000).to_csv('/Users/iouriolguin/Documents/test.csv')
#bankaya_merged.to_csv('/Users/iouriolguin/Documents/bankaya_merged.csv')
#bankaya_merged.to_parquet('/Users/iouriolguin/Documents/bankaya_merged.parquet')


# --- Identify all numerical variables in the dataset for the EDA analysis
numerical_vars = bankaya_eda.select_dtypes(include=['number']).columns.tolist()
numerical_vars = [var for var in numerical_vars if var not in ['customer_id', 'loan_id']] #---> remove id variables

# --- Identify all categorical variables in the dataset
categorical_vars =  bankaya_eda.select_dtypes(include=['object']).columns.tolist()
categorical_vars.remove("CDC_INQUIRY_ID")

# --- Identify the date columns 
date_vars = [col for col in bankaya_eda.columns if 'date' in col.lower() or 'datetime' in col.lower()]


# --- Univariate Analysis of the target variable

# Plotting the distribution of 'outcome_type'
fig = px.pie(bankaya_eda["target"], names='target', title='Distribution of Default')
fig.show()


# --- Create a histogram for certain variables
lst_histogram = ["max_days_late","account_to_application_days","MAX_CREDIT","TOTAL_PAYMENTS",'DELAYED_PAYMENTS','n_bnpl_apps','n_inquiries_l3m']
for col in lst_histogram:
    plot_histogram(bankaya_eda,col)


# Group by the application date and count the number of applications
application_time_series = bankaya_eda.groupby('APPLICATION_DATETIME').size().reset_index(name='Number_of_Applications')

# --- Time Series Plot for the most relevant KPIs
plot_time_series(application_time_series,'Number_of_Applications','APPLICATION_DATETIME','D','mean')
plot_time_series(bankaya_eda,'account_to_application_days','ACC_CREATION_DATETIME','D','sum')
plot_time_series(bankaya_eda,'n_sf_apps','LOAN_ORIGINATION_DATETIME','M','mean')
plot_time_series(bankaya_eda,'n_bnpl_apps','first_bnpl_app_date','M','mean')
plot_time_series(bankaya_eda,'n_inquiries_l3m','INQUIRY_DATE','M','mean')
plot_time_series(bankaya_eda,'TOTAL_PAYMENTS','LOAN_ORIGINATION_DATETIME','D','mean')


# --- Read the json file created in the prject in order to map and group all the Business types
# to reduce the cardinality in this variable

# Reading JSON from a file
with open("input/businees_type_catalog.json", 'r') as f:
    business_catalog_dict = json.load(f)

# Reverse the mapping to make it usable in `replace`
reverse_mapping = {k: old_k for old_k, old_v in business_catalog_dict.items() for k in old_v}

# Update column in the dataframe with the business categories
bankaya_eda['BUSINESS_TYPE'] = bankaya_eda['BUSINESS_TYPE'].replace(reverse_mapping)


# --- Pie Charts for Categorical Variables
# ---Remove the date features from the category variables
filtered_categorical_vars = list(set(categorical_vars) - set(date_vars))
filtered_categorical_vars = list(set(filtered_categorical_vars) - set(["CREDIT_TYPE","BUSINESS_TYPE","CURRENT_PAYMENT","PREVENTION_KEY"])) 

# --- Plot relevant category variables
for var in filtered_categorical_vars:
    plot_pie(bankaya_eda,var)


from ML_functions.plots import cross_values_plot


cross_values_plot(bankaya_eda,"max_days_late","BUSINESS_TYPE","ACCOUNT_TYPE",temp_col_name="days_late")


# --- Investigate the average max_days_late for different 'account_to_application_days' bins
bins = [0, 7, 30, 90, 180, 365, float('inf')]
labels = ['0-7 days', '8-30 days', '31-90 days', '91-180 days', '181-365 days', 'Over a year']
bankaya_eda['Application_Timeframe'] = pd.cut(bankaya_eda['account_to_application_days'], bins=bins, labels=labels)
# ---  Calculate the average max_days_late for each bin
avg_max_days_late = bankaya_eda.groupby('Application_Timeframe')['max_days_late'].mean().reset_index()

# -- Plot the results
fig = px.bar(avg_max_days_late, x='Application_Timeframe', y='max_days_late', title='Average max_days_late by Application Timeframe')
fig.show()


# Calculate the average max_days_late for each RESPONSABILITY_TYPE
avg_max_days_late_by_responsability = bankaya_eda.groupby('RESPONSABILITY_TYPE')['max_days_late'].mean().reset_index()

# Plot the results
fig = px.bar(avg_max_days_late_by_responsability, x='RESPONSABILITY_TYPE', y='max_days_late', title='Average max_days_late by RESPONSABILITY_TYPE')
fig.show()


from ML_functions.plots import plot_box, plot_cross_data


# Plot the distribution of CURRENT_BALANCE for different ACCOUNT_TYPEs
plot_box(bankaya_eda,'ACCOUNT_TYPE','max_days_late','Distribution of CURRENT_BALANCE by ACCOUNT_TYPE')


# --- Read the json file created in the prject in order to map and group all the Credit types
# to reduce the cardinality in this variable

# Reading JSON from a file
with open("input/credit_type.json", 'r') as f:
    credit_catalog_dict = json.load(f)

# Reverse the mapping to make it usable in `replace`
reverse_mapping = {k: old_k for old_k, old_v in credit_catalog_dict.items() for k in old_v}

# Update column in the dataframe with the business categories
bankaya_eda['sub_credit_type'] = bankaya_eda['CREDIT_TYPE'].replace(reverse_mapping)


plot_cross_data(bankaya_eda,"max_days_late","sub_credit_type","ACCOUNT_TYPE","days later","mean")


# --- Plot matrix correlation
matrix_correlation(bankaya_eda[numerical_vars],cols_to_exclude=None)


from ML_functions.util_functions import remove_highly_correlated_features


# --- Select the high correlated varoiables under a 0.7 threshold
high_correlated_lst = remove_highly_correlated_features(bankaya_eda[numerical_vars],threshold=0.7,cols_to_exclude=["target"])
high_correlated_lst

 === High correlated features to be removed from the Dataset====

['WORST_DELAY', 'BALANCE_DUE', 'BALANCE_DUE_WORST_DELAY']


# --- Select the the features to be used by removing high correlated variables
model_current_features = list(set(bankaya_merged.columns.to_list())- set(high_correlated_lst))
model_current_features = [var for var in model_current_features if 'id' not in var.lower() or "customer_id" in var]


# --- Create a new dataframe in order that not duplicate under the dataset
bankaya_model_df = bankaya_merged.copy()
bankaya_model_df = bankaya_model_df[model_current_features]


# --- Identify all numerical variables in the dataset for the modeling pprocesing. 
model_numerical_vars = bankaya_model_df.select_dtypes(include=['number']).columns.tolist() 

# --- Identify all categorical variables in the dataset
model_categorical_vars =  bankaya_model_df.select_dtypes(include=['object']).columns.tolist()

# --- Identify the date columns 
model_date_vars = [col for col in bankaya_model_df.columns if 'date' in col.lower() or 'datetime' in col.lower()]


# === Diagnostic outliers ===
for var in bankaya_model_df[model_numerical_vars]:
    if 'target' not in var and "customer_id" not in var:
        plot_diagnostic_outlier(bankaya_model_df[model_numerical_vars],var)


# --- Calculate the payment_to_due_ratio. This means how much of the current balance is covered by the payment amount. 
bankaya_model_df["payment_to_due_ratio"] = np.where(
    (bankaya_model_df["CURRENT_BALANCE"] == 0) & (bankaya_model_df["PAYMENT_AMOUNT"] == 0),
    0, round(bankaya_model_df["CURRENT_BALANCE"] / bankaya_model_df["PAYMENT_AMOUNT"],2))

# # --- Time to loan origination. It represents the duration in days between the application date and the 
# loan origination . 
bankaya_model_df["time_to_loan_days"] = -(bankaya_model_df["APPLICATION_DATETIME"] - bankaya_model_df["LOAN_ORIGINATION_DATETIME"]).dt.days

# --- Last payment days. A longer time might indicate financial distress or irregular payment behavior. 
bankaya_model_df["last_payment_days"]= abs(bankaya_model_df["LAST_PAYMENT_DATE"] - bankaya_model_df["LAST_PURCHASE_DATE"]).dt.days

# # Number of unique credit types by customer. Multiple types of credit could either indicate financial savviness or over-leveraging.
# #bankaya_model_df["n_credit_types"] = bankaya_model_df[]

# --- Outstanding balance. A higher outstanding balance relative to the credit limit could indicate financial distress.
bankaya_model_df["outstanding_balance"] = abs(bankaya_model_df["CREDIT_LIMIT"] - bankaya_model_df["CURRENT_BALANCE"])

# --- Loan life. The customer paid off the loan quickly, possibly indicating financial stability and low risk or 
# the loan had to be closed quickly due to default or restructuring, indicating high risk.
bankaya_model_df["loan_life_days"] = (bankaya_model_df["LOAN_CLOSING_DATE"] - bankaya_model_df["LOAN_OPENING_DATE"]).dt.days


# --- Read the json file created in the project in order to map and group all the Business types
# to reduce the cardinality in this variable

# Reading JSON from a file
with open("input/businees_type_catalog.json", 'r') as f:
    business_catalog_dict = json.load(f)

# Reverse the mapping to make it usable in `replace`
reverse_mapping = {k: old_k for old_k, old_v in business_catalog_dict.items() for k in old_v}

# Update column in the dataframe with the business categories
bankaya_model_df["gen_business_type"] = bankaya_model_df['BUSINESS_TYPE'].replace(reverse_mapping)


# --- Read the json file created in the prject in order to map and group all the Credit types
# to reduce the cardinality in this variable

# Reading JSON from a file
with open("input/credit_type.json", 'r') as f:
    credit_catalog_dict = json.load(f)

# Reverse the mapping to make it usable in `replace`
reverse_mapping = {k: old_k for old_k, old_v in credit_catalog_dict.items() for k in old_v}

# Update column in the dataframe with the business categories
bankaya_model_df['gen_credit_type'] = bankaya_model_df['CREDIT_TYPE'].replace(reverse_mapping)

# --- Reading JSON file from the prevention key scores

with open("input/prevention_key_scores.json","r") as f:
    risk_scores= json.load(f)


# --- Current features after the feature engenieering processing
cols_current_features =bankaya_model_df.columns.to_list()

# --- Remove unnecesary variables or varibales which weres used before. 
model_cols_to_remove = ["CURRENT_PAYMENT","TOTAL_REPORTED_PAYMENTS","CURRENT_PAYMENT",
                      "CREDIT_TYPE","BUSINESS_TYPE",'INQUIRY_TIME'] #+ model_date_vars

# --- Create the final list of variables to use for the model
model_remaining_features = list(set(cols_current_features) - set(model_cols_to_remove))


# --- Create the final datset which it will be grouped later to use for the model
bankaya_final_model_df = bankaya_model_df[model_remaining_features]

# ---- Lowercase all the columns in the datset
bankaya_final_model_df.columns = [col.lower() for col in bankaya_final_model_df.columns]
print(f"dataframe shape: {bankaya_final_model_df.shape}")
bankaya_final_model_df.head(5)

dataframe shape: (285237, 42)


# --- Identify all categorical variables after including the new categories for the Business Type 
model_categorical_vars =  bankaya_final_model_df.select_dtypes(include=['object']).columns.tolist()

# --- Identify all numerical variables in the dataset for the modeling pprocesing. 
model_numerical_vars = bankaya_final_model_df.select_dtypes(include=['number']).columns.tolist() 

# --- Select the OHE variables to be transformed
model_ohe_vars = ["account_type","currency","responsability_type"]

# --- Select the TE variables to be transformed
model_te_vars = ["gen_credit_type","gen_business_type","payment_frequency"]

print(f"Total variables so far: {len(model_categorical_vars)+len(model_numerical_vars)}")

Total variables so far: 27


# --- Create dummie variables or OHE for the categorical variables
bankaya_model_df_ohe = pd.get_dummies(bankaya_final_model_df, columns=model_ohe_vars)

# Show the new columns created as a result of One-Hot Encoding to verify
new_columns = [col for col in bankaya_model_df_ohe.columns if any(col.startswith(prefix) for prefix in model_ohe_vars)]
#new_columns[:10]  # Displaying only the first 10 for brevity


# Initialize the target encoder
target_encoder = ce.TargetEncoder(cols=model_te_vars)

# Fit and transform to get the target encoded columns
bankaya_final_model_df_encoded = target_encoder.fit_transform(
    bankaya_model_df_ohe[model_te_vars], 
    bankaya_model_df_ohe['target']
)

# Combine the encoded columns with the rest of your dataframe (excluding original categorical columns)
bankaya_final_model_df = pd.concat(
    [
        bankaya_model_df_ohe.drop(model_te_vars, axis=1), 
        bankaya_final_model_df_encoded
    ], 
    axis=1
)


# --- Create a dictionary for the aggregation to be done. 
aggregations = {}

# --- Loop through each column to decide the aggregation method
for col in bankaya_final_model_df.columns:
    if col == 'customer_id':
        continue  # Skip the customer_id column
    if bankaya_final_model_df[col].dtype == 'object':
        aggregations[col] = ['nunique']  # Number of unique categories for categorical features
    elif 'datetime' in str(bankaya_final_model_df[col].dtype):
        aggregations[col] = ['min', 'max']  # Min and Max for datetime features
    else:
        aggregations[col] = ['mean', 'sum', 'count']  # Mean, Sum, and Count for numerical features


# --- Mapping the prevention key with the scores given. 
bankaya_final_model_df['prevention_key_risk_score'] = bankaya_final_model_df['prevention_key'].map(risk_scores)

# --- Update the aggregations dictionary to include the new risk score column
aggregations['prevention_key_risk_score'] = ['mean', 'sum', 'count']


# Group by 'customer_id' and aggregate using the defined aggregations
grouped_df = bankaya_final_model_df.groupby('customer_id').agg(aggregations).reset_index()

# Flatten the MultiIndex in columns
grouped_df.columns = ['_'.join(col).strip() for col in grouped_df.columns.values]

# Rename the 'customer_id_' column to 'customer_id'
grouped_df.rename(columns={'customer_id_': 'customer_id'}, inplace=True)

# Display the first few rows of the grouped dataframe
grouped_df


#acc_creation_datetime_min	 - # --- Create a dictionary for the aggregation to be done. 
aggregations = {}

# --- Loop through each column to decide the aggregation method
for col in bankaya_final_model_df.columns:
    if col == 'customer_id':
        continue  # Skip the customer_id column
    if bankaya_final_model_df[col].dtype == 'object':
        aggregations[col] = ['nunique']  # Number of unique categories for categorical features
    elif 'datetime' in str(bankaya_final_model_df[col].dtype):
        aggregations[col] = ['min', 'max']  # Min and Max for datetime features
    else:
        aggregations[col] = ['mean', 'sum', 'count']  # Mean, Sum, and Count for numerical features


# --- Mapping the prevention key with the scores given. 
bankaya_final_model_df['prevention_key_risk_score'] = bankaya_final_model_df['prevention_key'].map(risk_scores)

# --- Update the aggregations dictionary to include the new risk score column
aggregations['prevention_key_risk_score'] = ['mean', 'sum', 'count']


# Group by 'customer_id' and aggregate using the defined aggregations
grouped_df = bankaya_final_model_df.groupby('customer_id').agg(aggregations).reset_index()

# Flatten the MultiIndex in columns
grouped_df.columns = ['_'.join(col).strip() for col in grouped_df.columns.values]

# Rename the 'customer_id_' column to 'customer_id'
grouped_df.rename(columns={'customer_id_': 'customer_id'}, inplace=True)

# Display the first few rows of the grouped dataframe
grouped_df


#acc_creation_datetime_min	 - # --- Create a dictionary for the aggregation to be done. 
aggregations = {}

# --- Loop through each column to decide the aggregation method
for col in bankaya_final_model_df.columns:
    if col == 'customer_id':
        continue  # Skip the customer_id column
    if bankaya_final_model_df[col].dtype == 'object':
        aggregations[col] = ['nunique']  # Number of unique categories for categorical features
    elif 'datetime' in str(bankaya_final_model_df[col].dtype):
        aggregations[col] = ['min', 'max']  # Min and Max for datetime features
    else:
        aggregations[col] = ['mean', 'sum', 'count']  # Mean, Sum, and Count for numerical features


# --- Mapping the prevention key with the scores given. 
bankaya_final_model_df['prevention_key_risk_score'] = bankaya_final_model_df['prevention_key'].map(risk_scores)

# --- Update the aggregations dictionary to include the new risk score column
aggregations['prevention_key_risk_score'] = ['mean', 'sum', 'count']


# Group by 'customer_id' and aggregate using the defined aggregations
grouped_df = bankaya_final_model_df.groupby('customer_id').agg(aggregations).reset_index()

# Flatten the MultiIndex in columns
grouped_df.columns = ['_'.join(col).strip() for col in grouped_df.columns.values]

# Rename the 'customer_id_' column to 'customer_id'
grouped_df.rename(columns={'customer_id_': 'customer_id'}, inplace=True)

# --- Create customer lifetime feature

grouped_df['customer_lifetime'] = abs(grouped_df["acc_creation_datetime_min"] - grouped_df["application_datetime_max"]).dt.days

# Display the first few rows of the grouped dataframe
grouped_df


# customer_lifespin = acc_creation_datetime_min	 - application_datetime_max


# --- Choose only the variables I use for the model

model_features_final = [
    'customer_id','target_mean','customer_lifetime','n_sf_apps_sum','prevention_key_nunique',
    'account_to_application_days_count', 'loan_life_days_mean', 'max_credit_sum',
    'total_payments_sum', 'n_inquiries_l3m_mean', 'payment_amount_sum',
    'payment_to_due_ratio_mean', 'time_to_loan_days_mean',
    'credit_limit_sum', 'n_bnpl_approved_apps_mean', 'outstanding_balance_mean',
    'n_inquiries_l6m_mean', 'n_bnpl_apps_sum', 'max_days_late_sum', 'last_payment_days_sum',
    'delayed_payments_sum', 'current_balance_sum', 'account_type_CREDITO DE HABILITACION O AVIO_sum',
    'account_type_HIPOTECA_sum', 'account_type_PAGOS FIJOS_sum', 'account_type_POR DETERMINAR_sum',
    'account_type_REVOLVENTE_sum', 'account_type_SIN LIMITE PREESTABLECIDO_sum', 'currency_DOLARES_sum',
    'currency_PESOS MEXICANOS_sum', 'currency_UNIDADES DE INVERSION_sum', 'responsability_type_AVAL_sum',
    'responsability_type_INDIVIDUAL (TITULAR)_sum', 'responsability_type_MANCOMUNADO_sum',
    'responsability_type_OBLIGATORIO SOLIDARIO_sum', 'responsability_type_TITULAR CON AVAL_sum',
    'prevention_key_risk_score_mean','gen_credit_type_mean','gen_business_type_mean','payment_frequency_mean'
]


# --- Create the final dataframe for the model
bankaya_grouped_df = grouped_df[model_features_final]

# ---- Lowercase all the columns in the datset
bankaya_grouped_df.columns = [col.lower() for col in bankaya_grouped_df.columns ]


# ---- Rename the columns for the datset and get a better understanding
new_cols_named = {
    'target_mean':'target','n_sf_apps_sum':'n_sf_apps',
    'account_to_application_days_count' : 'total_credits','max_credit_sum':'max_credit', 'total_payments_sum': 'total_payments_sum',
    'payment_amount_sum':'payment_amount','payment_to_due_ratio_mean':'payment_to_due_ratio','credit_limit_sum':'credit_limit','outstanding_balance_mean':'outstanding_balance',
    'n_bnpl_apps_sum':'n_bnpl_apps','max_days_late_sum':'max_days_late','last_payment_days_sum':'last_payment_days','delayed_payments_sum':'delayed_payments',
    'current_balance_sum':'current_balance','account_type_CREDITO DE HABILITACION O AVIO_sum':'account_type_CREDITO_DE_HABILITACION_O_AVIO',
    'account_type_HIPOTECA_sum':'account_type_HIPOTECA','account_type_PAGOS FIJOS_sum':'account_type_PAGOS_FIJOS',
    'account_type_POR DETERMINAR_sum':'account_type_POR_DETERMINAR','account_type_REVOLVENTE_sum':'account_type_REVOLVENTE',
    'account_type_SIN LIMITE PREESTABLECIDO_sum': 'account_type_SIN_LIMITE_PREESTABLECIDO','currency_DOLARES_sum':'currency_DOLARES',
    'currency_PESOS MEXICANOS_sum':'currency_PESOS_MEXICANOS','currency_UNIDADES DE INVERSION_sum':'currency_UNIDADES_DE_INVERSION',
    'responsability_type_AVAL_sum':'responsability_type_AVAL','responsability_type_INDIVIDUAL (TITULAR)_sum':'responsability_type_INDIVIDUAL(TITULAR)',
    'responsability_type_MANCOMUNADO_sum':'responsability_type_MANCOMUNADO', 'responsability_type_OBLIGATORIO SOLIDARIO_sum': 'responsability_type_OBLIGATORIO_SOLIDARIO',
    'responsability_type_TITULAR CON AVAL_sum':'responsability_type_TITULAR_CON_AVAL','prevention_key_risk_score_mean':'prevention_key_risk_score',
    'gen_credit_type_mean':'gen_credit_type','gen_business_type_mean':'gen_business_type','payment_frequency_mean':'payment_frequency'
}

# Use the rename method with the dictionary to rename the columns
bankaya_grouped_df = bankaya_grouped_df.rename(columns=new_cols_named)

# --- Display final dataset t be modeled 
bankaya_grouped_df


from sklearn.cluster import KMeans
import numpy as np


# Prepare the data for clustering
# Remove NaN values and reshape the data to meet the requirements of the KMeans algorithm
customer_tenure_data = bankaya_grouped_df['customer_lifetime'].dropna().values.reshape(-1, 1)

# Using the elbow method to find the optimal number of clusters
inertia = []
for i in range(1, 11):
    kmeans = KMeans(n_clusters=i, random_state=0)
    kmeans.fit(customer_tenure_data)
    inertia.append(kmeans.inertia_)

# Plotting the elbow graph
plt.figure(figsize=(10, 5))
plt.plot(range(1, 11), inertia, marker='o')
plt.title('Elbow Method For Optimal Number of Clusters')
plt.xlabel('Number of Clusters')
plt.ylabel('Inertia')
plt.show()


# ---Perform K-means clustering with 3 clusters (as one of the potential elbow points)
kmeans = KMeans(n_clusters=3, random_state=0)
bankaya_grouped_df['Lifetime_Cluster'] = kmeans.fit_predict(customer_tenure_data)

#---Calculate the mean and count for each cluster
cluster_summary = bankaya_grouped_df.groupby('Lifetime_Cluster')['customer_lifetime'].agg(['mean', 'count']).reset_index()

# ---Sort clusters by the mean tenure
cluster_summary = cluster_summary.sort_values(by='mean')

# ---Rename clusters based on the sorted mean tenure
cluster_map = {row['Lifetime_Cluster']: 'Cluster_{}'.format(index) for index, row in cluster_summary.iterrows()}
bankaya_grouped_df['Lifetime_Cluster'] = bankaya_grouped_df['Lifetime_Cluster'].map(cluster_map)
bankaya_grouped_df['Lifetime_Cluster'] = bankaya_grouped_df['Lifetime_Cluster'].replace({'Cluster_1': 'Long_Tenure',
                                                                                             'Cluster_0': 'Medium_Tenure',
                                                                                             'Cluster_2': 'Short_Tenure'})


# ---Show the cluster summary and some sample rows
print(cluster_summary)

bankaya_grouped_df.sample(5)

   Lifetime_Cluster        mean  count
1                 1   70.579334   4355
0                 0  261.493224   3173
2                 2  516.891342   1721


# Calculate Recency, Frequency, and Monetary value for each customer
# Recency
recency_df = bankaya_final_model_df.groupby('customer_id')['last_payment_date'].max().reset_index()
recency_df.columns = ['customer_id', 'Recent_Payment_Date']
recency_df['Recency'] = (bankaya_final_model_df['last_payment_date'].max() - recency_df['Recent_Payment_Date']).dt.days

# Frequency
frequency_df = bankaya_final_model_df.groupby('customer_id')['application_datetime'].count().reset_index()
frequency_df.columns = ['customer_id', 'Frequency']

# Monetary
monetary_df = bankaya_final_model_df.groupby('customer_id')['current_balance'].mean().reset_index()
monetary_df.columns = ['customer_id', 'Monetary']

# Merge these dataframes to create the RFM table
rfm_df = pd.merge(recency_df, frequency_df, on='customer_id')
rfm_df = pd.merge(rfm_df, monetary_df, on='customer_id')

# Show some statistics and sample rows for the RFM features
rfm_stats = rfm_df[['Recency', 'Frequency', 'Monetary']].describe()
rfm_sample = rfm_df.sample(5)

rfm_stats


rfm_sample


# --- Further segmenting the RFM scores using quantiles
# For Recency, lower the value, better the segment. Hence, labels are in reverse order
rfm_df['R_Segment'] = pd.qcut(rfm_df['Recency'], 4, labels=['4', '3', '2', '1'])

# --- For Frequency and Monetary, higher the value, better the segment
rfm_df['F_Segment'] = pd.qcut(rfm_df['Frequency'], 4, labels=['1', '2', '3', '4'])
rfm_df['M_Segment'] = pd.qcut(rfm_df['Monetary'], 4, labels=['1', '2', '3', '4'])

# --- Combine the individual segments to create a combined RFM score
rfm_df['RFM_Score'] = rfm_df['R_Segment'].astype(str) + rfm_df['F_Segment'].astype(str) + rfm_df['M_Segment'].astype(str)

# --- -Define the RFM level function
def rfm_level(df):
    if df['RFM_Score'] == '444':
        return 'Champion'
    elif df['F_Segment'] == '4' and df['M_Segment'] == '4':
        return 'Loyal Customer'
    elif df['R_Segment'] == '1':
        return 'At Risk'
    elif df['R_Segment'] == '2' and df['F_Segment'] == '1':
        return 'Needs Attention'
    else:
        return 'Average'

# --- Calculate RFM Level
rfm_df['RFM_Level'] = rfm_df.apply(rfm_level, axis=1)

# --- Show the distribution of RFM Level and some sample rows
rfm_level_summary = rfm_df['RFM_Level'].value_counts().reset_index()
rfm_level_summary.columns = ['RFM_Level', 'Count']
rfm_sample_segmented = rfm_df

rfm_level_summary


# --- Display the RFM scores. I will use these scores to included in the model
rfm_sample_segmented


## --- Enhancing the DataFrame with RFM Scores for Modeling:
bankaya_model_merged_df = bankaya_grouped_df.merge(rfm_df[['customer_id', 'RFM_Score']], on='customer_id', how='left')

# Replace 'nan' strings with actual NaN values (np.nan)
bankaya_model_merged_df['RFM_Score'] = bankaya_model_merged_df['RFM_Score'].replace('nan', np.nan, regex=True)

# Now I can fill NaN values with a placeholder, or drop them, etc.
bankaya_model_merged_df['RFM_Score'] = bankaya_model_merged_df['RFM_Score'].fillna(0)

# Attempt conversion to integer
bankaya_model_merged_df['RFM_Score'] = bankaya_model_merged_df['RFM_Score'].astype(int)

bankaya_model_merged_df.sample(5)


# --- Manual mappings to assign the values to each category
lifetime_cluster_mapping = {
    'Short_Tenure': 1,
    'Medium_Tenure': 2,
    'Long_Tenure': 3,
}
# --- Mapping the lifetime cluster and the RFM importance with the scores given. 
bankaya_model_merged_df['Lifetime_Cluster_score'] = bankaya_model_merged_df['Lifetime_Cluster'].map(lifetime_cluster_mapping)

# --- Drop unnecessary columns 
bankaya_model_merged_df.drop("Lifetime_Cluster",inplace=True,axis=1)


# <<<< Function to create the pipeline for the model >>>>
def transform_data_test():

    continuous_pipeline = Pipeline([
        ("imputer",SimpleImputer(strategy="mean")),
        ("scaler",RobustScaler(with_centering=True,with_scaling=True))

    ])

    preproccessing_pipeline = ColumnTransformer([
        ("continous",continuous_pipeline,model_numerical_vars)

    ],
        remainder = "drop"
    )

    pipeline = ImbPipeline(steps=([
        ("preprocess",preproccessing_pipeline),
        ("smote", SMOTE(random_state=42)) 
        
        
    ]))
    return pipeline


#!pip install imbalanced-learn


# --- Identify all numerical variables in the dataset for the modeling pprocesing. 
model_numerical_vars = [col for col in bankaya_model_merged_df.columns if 'target'not in col and 'customer_id' not in col ]


# --- Pipeline process to be applied to the X_train and X_test
pipeline_data = transform_data_test()
pipeline_data

Pipeline(steps=[('preprocess',
                 ColumnTransformer(transformers=[('continous',
                                                  Pipeline(steps=[('imputer',
                                                                   SimpleImputer()),
                                                                  ('scaler',
                                                                   RobustScaler())]),
                                                  ['customer_lifetime',
                                                   'n_sf_apps',
                                                   'prevention_key_nunique',
                                                   'total_credits',
                                                   'loan_life_days_mean',
                                                   'max_credit',
                                                   'total_payments_sum',
                                                   'n_inquiries_l3m_mean',
                                                   'payment_amount',
                                                   'payment_to_due_ratio',
                                                   'time_to_l...
                                                   'habilitacion o avio_sum',
                                                   'account_type_hipoteca_sum',
                                                   'account_type_pagos '
                                                   'fijos_sum',
                                                   'account_type_por '
                                                   'determinar_sum',
                                                   'account_type_revolvente_sum',
                                                   'account_type_sin limite '
                                                   'preestablecido_sum',
                                                   'currency_dolares_sum',
                                                   'currency_pesos '
                                                   'mexicanos_sum',
                                                   'currency_unidades de '
                                                   'inversion_sum',
                                                   'responsability_type_aval_sum', ...])])),
                ('smote', SMOTE(random_state=42))])

Pipeline(steps=[('preprocess',
                 ColumnTransformer(transformers=[('continous',
                                                  Pipeline(steps=[('imputer',
                                                                   SimpleImputer()),
                                                                  ('scaler',
                                                                   RobustScaler())]),
                                                  ['customer_lifetime',
                                                   'n_sf_apps',
                                                   'prevention_key_nunique',
                                                   'total_credits',
                                                   'loan_life_days_mean',
                                                   'max_credit',
                                                   'total_payments_sum',
                                                   'n_inquiries_l3m_mean',
                                                   'payment_amount',
                                                   'payment_to_due_ratio',
                                                   'time_to_l...
                                                   'habilitacion o avio_sum',
                                                   'account_type_hipoteca_sum',
                                                   'account_type_pagos '
                                                   'fijos_sum',
                                                   'account_type_por '
                                                   'determinar_sum',
                                                   'account_type_revolvente_sum',
                                                   'account_type_sin limite '
                                                   'preestablecido_sum',
                                                   'currency_dolares_sum',
                                                   'currency_pesos '
                                                   'mexicanos_sum',
                                                   'currency_unidades de '
                                                   'inversion_sum',
                                                   'responsability_type_aval_sum', ...])])),
                ('smote', SMOTE(random_state=42))])

ColumnTransformer(transformers=[('continous',
                                 Pipeline(steps=[('imputer', SimpleImputer()),
                                                 ('scaler', RobustScaler())]),
                                 ['customer_lifetime', 'n_sf_apps',
                                  'prevention_key_nunique', 'total_credits',
                                  'loan_life_days_mean', 'max_credit',
                                  'total_payments_sum', 'n_inquiries_l3m_mean',
                                  'payment_amount', 'payment_to_due_ratio',
                                  'time_to_loan_days_mean', 'credit_limit',...
                                  'current_balance',
                                  'account_type_credito de habilitacion o '
                                  'avio_sum',
                                  'account_type_hipoteca_sum',
                                  'account_type_pagos fijos_sum',
                                  'account_type_por determinar_sum',
                                  'account_type_revolvente_sum',
                                  'account_type_sin limite preestablecido_sum',
                                  'currency_dolares_sum',
                                  'currency_pesos mexicanos_sum',
                                  'currency_unidades de inversion_sum',
                                  'responsability_type_aval_sum', ...])])

['customer_lifetime', 'n_sf_apps', 'prevention_key_nunique', 'total_credits', 'loan_life_days_mean', 'max_credit', 'total_payments_sum', 'n_inquiries_l3m_mean', 'payment_amount', 'payment_to_due_ratio', 'time_to_loan_days_mean', 'credit_limit', 'n_bnpl_approved_apps_mean', 'outstanding_balance', 'n_inquiries_l6m_mean', 'n_bnpl_apps', 'max_days_late', 'last_payment_days', 'delayed_payments', 'current_balance', 'account_type_credito de habilitacion o avio_sum', 'account_type_hipoteca_sum', 'account_type_pagos fijos_sum', 'account_type_por determinar_sum', 'account_type_revolvente_sum', 'account_type_sin limite preestablecido_sum', 'currency_dolares_sum', 'currency_pesos mexicanos_sum', 'currency_unidades de inversion_sum', 'responsability_type_aval_sum', 'responsability_type_individual (titular)_sum', 'responsability_type_mancomunado_sum', 'responsability_type_obligatorio solidario_sum', 'responsability_type_titular con aval_sum', 'prevention_key_risk_score', 'gen_credit_type', 'gen_business_type', 'payment_frequency', 'RFM_Score', 'Lifetime_Cluster_score']

SimpleImputer()

RobustScaler()

SMOTE(random_state=42)


# <<<< Búsqueda de hiperparametros de cada modelo >>>>

classifiers = {
    
    'Bagging': BaggingClassifier(random_state=314),
    'RF': RandomForestClassifier(random_state=314,n_jobs=-1),
    'AdaBoost': AdaBoostClassifier(base_estimator = DecisionTreeClassifier(max_depth=2, random_state=314),random_state=314),
    'GBoosting': GradientBoostingClassifier(random_state=314),
    "XGBoost": xgb.XGBClassifier(random_state = 314),
    'Logistic_r':LogisticRegression(random_state = 314),
    'D_tree':DecisionTreeClassifier(criterion='gini', random_state=314)
}

params = {
          
            'RF':{'n_estimators':range(1,11),
                  'max_depth': range(1,11),
                  "min_samples_split":[2,4]
                 },
    
            'AdaBoost': {'learning_rate':[1e-5, 1e-4,1e-3, 1e-2, 1e-1, 1.0],
                         'n_estimators':range(1,11)
                         },
    
            'GBoosting': {'max_depth':range(1,11),
                          'n_estimators':range(1,11),
                          'learning_rate':[1e-3, 1e-2, 1e-1,1.0]
                         },
    
            'XGBoost': {'max_depth':range(1,11),
                        'n_estimators':range(1,11),
                        "learning_rate":[1e-3, 1e-2, 1e-1,1.0],
                        'eval_metric': ['logloss']
                       },
    
            'Logistic_r': {'penalty' : ['l1'],
                           'C': [0.01, 0.03,0.04,0.05,0.06],
                           'solver': ['saga'],
                           "max_iter":[5000,10000]
                           },
    
            'D_tree': { 'max_depth': [2,4,5,6,7,8],
                        'min_samples_leaf': [1, 5, 10, 20, 35],
                        'min_samples_split': [2, 15, 30, 60],
                       }
        }


#--- Split DataFrame into train and test
y_bankaya_model = bankaya_model_merged_df["target"]
X_bankaya_model = bankaya_model_merged_df.drop(["target","customer_id"],axis=1) # --- Remove unneccessary columns


# --- Split model into Train and Test
X_train, X_test, y_train, y_test = train_test_split(X_bankaya_model,y_bankaya_model, test_size=0.2, random_state=42, stratify=y_bankaya_model)

# Replace infinite values with NaN
X_train.replace([np.inf, -np.inf], np.nan, inplace=True)
X_test.replace([np.inf, -np.inf], np.nan, inplace=True)


print("Number transactions X_train dataset: ", X_train.shape)
print("Number transactions y_train dataset: ", y_train.shape)
print("Number transactions X_test dataset: ", X_test.shape)
print("Number transactions y_test dataset: ", y_test.shape)
print("\n========Variables objetivo dataset Train===========\n")
print("Before OverSampling, counts of target '1': {}".format(sum(y_train==1)))
print("Before OverSampling, counts of target '0': {} \n".format(sum(y_train==0)))

Number transactions X_train dataset:  (7399, 40)
Number transactions y_train dataset:  (7399,)
Number transactions X_test dataset:  (1850, 40)
Number transactions y_test dataset:  (1850,)

========Variables objetivo dataset Train===========

Before OverSampling, counts of target '1': 1380
Before OverSampling, counts of target '0': 6019


# --- Use fit_resample method just in train to apply the resample SMOTE method
X_train_resampled, y_train_resampled = pipeline_data.fit_resample(X_train, y_train) 

# First, get the preprocessing and transformation steps without the SMOTE step
preprocess_pipeline = pipeline_data[:-1]

# --- Apply the transform method to the test set using the obtained preprocessing pipeline
X_test_scaled = preprocess_pipeline.transform(X_test) # apply transform just in test


# Convert the scaled and transformed X_train back to a DataFrame
X_train_resampled_df = pd.DataFrame(X_train_resampled, columns=model_numerical_vars)

# Convert the scaled and transformed X_test back to a DataFrame
X_test_scaled_df = pd.DataFrame(X_test_scaled, columns=model_numerical_vars)


print('After SMOTE, the shape of X_train_under: {}'.format(X_train_resampled.shape))
print('After SMOTE, the shape of y_train_under: {} \n'.format(y_train_resampled.shape))

print("After SMOTE, counts of label '1': {}".format(sum(y_train_resampled==1)))
print("After SMOTE, counts of label '0': {}".format(sum(y_train_resampled==0)))

After SMOTE, the shape of X_train_under: (12038, 40)
After SMOTE, the shape of y_train_under: (12038,) 

After SMOTE, counts of label '1': 6019
After SMOTE, counts of label '0': 6019


import warnings
warnings.filterwarnings("ignore")
from sklearn.experimental import enable_halving_search_cv  
from sklearn.model_selection import HalvingGridSearchCV
from sklearn.experimental import enable_halving_search_cv


# Create an empety dictionary where the model will be stored
clasifier_dict = {}


# < ===== GridSeacrh del modelo XGBoost ====== >

for name in list(params.keys()):
    if name == "XGBoost":
        print("================>")
        print(name)
        print("================>")
        
        clasifier_dict[name] = GridSearchCV(classifiers[name],
                                            params[name],
                                            #param_grid= pipeline["classifier"].available_models[name],
                                            scoring='roc_auc', 
                                            return_train_score=True, 
                                            n_jobs=-1, cv=3, verbose=1)
        
   
        
        clasifier_dict[name].fit(X_train_resampled, y_train_resampled)

================>
XGBoost
================>
Fitting 3 folds for each of 400 candidates, totalling 1200 fits


# < ======= GridSeacrh del modelo Logistic regression ========= >

for name in list(params.keys()):
    if name == "Logistic_r":
        print("================>")
        print(name)
        print("================>")
        clasifier_dict[name] = GridSearchCV(estimator= classifiers[name],
                                            param_grid=params[name],
                                            #param_grid= pipeline["classifier"].available_models[name],
                                            scoring="roc_auc", 
                                            return_train_score=True, 
                                            n_jobs=-1, cv=3, verbose=1)
        
        clasifier_dict[name].fit(X_train_resampled, y_train_resampled)

================>
Logistic_r
================>
Fitting 3 folds for each of 10 candidates, totalling 30 fits


# < ======= GridSeacrh del modelo GBoosting ========= >

for name in list(params.keys()):
    if name == "GBoosting":
        print("================>")
        print(name)
        print("================>")
        clasifier_dict[name] = GridSearchCV(estimator= classifiers[name],
                                            param_grid=params[name],
                                            #param_grid= pipeline["classifier"].available_models[name],
                                            scoring="roc_auc", 
                                            return_train_score=True, 
                                            n_jobs=-1, cv=3, verbose=1)
        
        clasifier_dict[name].fit(X_train_resampled, y_train_resampled)

================>
GBoosting
================>
Fitting 3 folds for each of 400 candidates, totalling 1200 fits


# < ======= GridSeacrh del modelo AdaBoost ========= >

for name in list(params.keys()):
    if name == "AdaBoost":
        print("================>")
        print(name)
        print("================>")
        clasifier_dict[name] = GridSearchCV(estimator= classifiers[name],
                                            param_grid=params[name],
                                            scoring='roc_auc', 
                                            return_train_score=True, 
                                            n_jobs=-1, cv=3, verbose=1)
        
        clasifier_dict[name].fit(X_train_resampled, y_train_resampled)


# < ======= GridSeacrh del modelo Random Forest ========= >

for name in list(params.keys()):
    if name == "RF":
        print("================>")
        print(name)
        print("================>")
        clasifier_dict[name] = GridSearchCV(estimator= classifiers[name],
                                            param_grid=params[name],
                                            scoring='roc_auc', 
                                            return_train_score=True, 
                                            n_jobs=-1, cv=3, verbose=1)
        
        clasifier_dict[name].fit(X_train_resampled, y_train_resampled)

================>
RF
================>
Fitting 3 folds for each of 200 candidates, totalling 600 fits


# < ======= GridSeacrh del modelo Decision Tree ========= >

for name in list(params.keys()):
    if name == "D_tree":
        print("================>")
        print(name)
        print("================>")
        clasifier_dict[name] = GridSearchCV(estimator= classifiers[name],
                                            param_grid=params[name],
                                            scoring='roc_auc', 
                                            return_train_score=True, 
                                            n_jobs=-1, cv=3, verbose=1)
        
        clasifier_dict[name].fit(X_train_resampled, y_train_resampled)

================>
D_tree
================>
Fitting 3 folds for each of 120 candidates, totalling 360 fits


clasifier_dict

{'XGBoost': GridSearchCV(cv=3,
              estimator=XGBClassifier(base_score=None, booster=None,
                                      callbacks=None, colsample_bylevel=None,
                                      colsample_bynode=None,
                                      colsample_bytree=None, device=None,
                                      early_stopping_rounds=None,
                                      enable_categorical=False, eval_metric=None,
                                      feature_types=None, gamma=None,
                                      grow_policy=None, importance_type=None,
                                      interaction_constraints=None,
                                      learning_rate=None,...
                                      max_leaves=None, min_child_weight=None,
                                      missing=nan, monotone_constraints=None,
                                      multi_strategy=None, n_estimators=None,
                                      n_jobs=None, num_parallel_tree=None,
                                      random_state=314, ...),
              n_jobs=-1,
              param_grid={'eval_metric': ['logloss'],
                          'learning_rate': [0.001, 0.01, 0.1, 1.0],
                          'max_depth': range(1, 11),
                          'n_estimators': range(1, 11)},
              return_train_score=True, scoring='roc_auc', verbose=1),
 'Logistic_r': GridSearchCV(cv=3, estimator=LogisticRegression(random_state=314), n_jobs=-1,
              param_grid={'C': [0.01, 0.03, 0.04, 0.05, 0.06],
                          'max_iter': [5000, 10000], 'penalty': ['l1'],
                          'solver': ['saga']},
              return_train_score=True, scoring='roc_auc', verbose=1),
 'GBoosting': GridSearchCV(cv=3, estimator=GradientBoostingClassifier(random_state=314),
              n_jobs=-1,
              param_grid={'learning_rate': [0.001, 0.01, 0.1, 1.0],
                          'max_depth': range(1, 11),
                          'n_estimators': range(1, 11)},
              return_train_score=True, scoring='roc_auc', verbose=1),
 'AdaBoost': GridSearchCV(cv=3,
              estimator=AdaBoostClassifier(base_estimator=DecisionTreeClassifier(max_depth=2,
                                                                                 random_state=314),
                                           random_state=314),
              n_jobs=-1,
              param_grid={'learning_rate': [1e-05, 0.0001, 0.001, 0.01, 0.1,
                                            1.0],
                          'n_estimators': range(1, 11)},
              return_train_score=True, scoring='roc_auc', verbose=1),
 'RF': GridSearchCV(cv=3,
              estimator=RandomForestClassifier(n_jobs=-1, random_state=314),
              n_jobs=-1,
              param_grid={'max_depth': range(1, 11), 'min_samples_split': [2, 4],
                          'n_estimators': range(1, 11)},
              return_train_score=True, scoring='roc_auc', verbose=1),
 'D_tree': GridSearchCV(cv=3, estimator=DecisionTreeClassifier(random_state=314),
              n_jobs=-1,
              param_grid={'max_depth': [2, 4, 5, 6, 7, 8],
                          'min_samples_leaf': [1, 5, 10, 20, 35],
                          'min_samples_split': [2, 15, 30, 60]},
              return_train_score=True, scoring='roc_auc', verbose=1)}


# ======= Results of the Model  ==============
test_scores = []

for name in list(params.keys()):
    try:
        print(name)
        print("===========>")
        print("Number of configurations: {}\n".format(pd.DataFrame.from_dict(clasifier_dict[name].cv_results_).shape[0]))
        aux = pd.DataFrame.from_dict(clasifier_dict[name].cv_results_).sort_values(by="mean_test_score",ascending=False).iloc[0]
        roc_cv_test = aux['mean_test_score']
        roc_cv_train = aux['mean_train_score']

        roc_train = roc_auc_score(y_train_resampled, clasifier_dict[name].predict(X_train_resampled))
        roc_test = roc_auc_score(y_test, clasifier_dict[name].predict(X_test_scaled))

        test_scores.append((name,roc_train,roc_cv_train,roc_cv_test,roc_test))
        
    except:
        print("Model {} is not store in the database\n".format(name))

RF
===========>
Number of configurations: 200

AdaBoost
===========>
Number of configurations: 60

GBoosting
===========>
Number of configurations: 400

XGBoost
===========>
Number of configurations: 400

Logistic_r
===========>
Number of configurations: 10

D_tree
===========>
Number of configurations: 120


#Performance of the Models

resultado = pd.DataFrame(test_scores,columns=['Classifier','AUC_train','AUC_CV_train','AUC_validation','AUC_test'])

print("\n=========== Performance Table ============")
display(resultado)

best_clas = resultado.sort_values(by="AUC_validation",ascending=False).iloc[0]['Classifier']

print("El mejor clasificador es: {}".format(best_clas))

=========== Performance Table ============

El mejor clasificador es: XGBoost


# --- Get the best model
best_model = clasifier_dict["XGBoost"].best_estimator_

# --- Get feature importances from the best model
feature_importances = best_model.feature_importances_


# >>>> Confusion Matrix XGBoost <<<<

y_target = y_test
y_predicted = clasifier_dict["XGBoost"].predict(X_test_scaled)

cm = confusion_matrix(y_target,
                      y_predicted)

fig, ax = plot_confusion_matrix(conf_mat=cm)
plt.show()

precision = precision_score(y_target, y_predicted)
recall = recall_score(y_target, y_predicted)
F1=f1_score(y_target, y_predicted)

print('Precision: {:.2f} \nRecall {:.2f}\nF1 score {:.2f}'.format(precision *100, recall * 100,F1 * 100 ))

Precision: 98.85 
Recall 100.00
F1 score 99.42


import pandas as pd
import plotly.express as px

# I'm assuming 'model_numerical_vars' and 'feature_importances' are defined elsewhere in your code
df_importance = pd.DataFrame({
    'Feature': model_numerical_vars,
    'Importance': feature_importances
})

df_importance = df_importance.sort_values('Importance', ascending=True)

fig = px.bar(df_importance, x='Importance', y='Feature', orientation='h', 
             title='Feature Importance XGBoost', labels={'Importance': 'Importance Score'})

# Set the figure size
fig.update_layout(
    autosize=False,
    width=1000,    # You can adjust the width to your preference
    height=800,    # You can adjust the height to make sure all labels are visible
    margin=dict(
        l=50,  # left margin, increase if your y labels are getting cut off
        r=50,  # right margin
        b=100, # bottom margin
        t=100, # top margin
    )
)

# You can also modify the font size of the labels if necessary
fig.update_layout(
    title_font_size=20,
    font=dict(
        family="Arial, Balto, Courier New, Droid Sans",  # Choose a font family
        size=12,  # Set the global font size
        color="RebeccaPurple"  # Set the font color
    )
)

	customer_id	loan_id	ACC_CREATION_DATETIME	APPLICATION_DATETIME	LOAN_ORIGINATION_DATETIME	max_days_late	account_to_application_days	n_sf_apps	first_app_date	last_app_date	n_bnpl_apps	n_bnpl_approved_apps	first_bnpl_app_date	last_bnpl_app_date	n_inquiries_l3m	n_inquiries_l6m
0	1223	1	2021-08-23 13:57:56	2022-04-26 07:00:00	2022-10-20 18:26:35	5	245	1.0	2022-04-25	2022-04-25	1.0	0.0	2022-04-26 02:15:17.742	2022-04-26 02:15:17.742	91.0	91.0
1	5190	2	2022-04-26 09:57:25	2022-04-26 07:00:00	2022-12-22 18:03:32	2	0	NaN	NaT	NaT	NaN	NaN	NaT	NaT	NaN	NaN
2	5194	3	2022-04-26 12:22:35	2022-04-26 07:00:00	2022-08-09 17:31:05	4	0	NaN	NaT	NaT	NaN	NaN	NaT	NaT	NaN	NaN
3	3978	4	2022-03-09 11:26:55	2022-04-26 07:00:00	2022-11-11 19:48:08	2	47	1.0	2022-03-17	2022-03-17	2.0	0.0	2022-03-19 22:02:50.555	2022-03-09 16:41:49.675	13.0	13.0
4	4535	5	2022-04-01 14:28:42	2022-04-26 07:00:00	2022-07-24 12:07:15	2	24	1.0	2022-04-01	2022-04-01	1.0	0.0	2022-04-01 19:37:56.185	2022-04-01 19:37:56.185	21.0	21.0
5	3604	6	2022-02-21 11:55:32	2022-05-05 07:00:00	2022-12-08 07:12:18	7	72	1.0	2022-02-21	2022-02-21	NaN	NaN	NaT	NaT	26.0	26.0
6	271	7	2021-04-06 16:30:21	2022-05-05 07:00:00	2023-04-21 15:40:10	0	393	NaN	NaT	NaT	1.0	0.0	2022-05-04 23:18:03.604	2022-05-04 23:18:03.604	11.0	11.0

	customer_id	INQUIRY_TIME	CDC_INQUIRY_ID	INQUIRY_DATE	PREVENTION_KEY	CURRENCY	MAX_CREDIT	CREDIT_LIMIT	PAYMENT_AMOUNT	UPDATE_DATE	LOAN_OPENING_DATE	LOAN_CLOSING_DATE	WORST_DELAY_DATE	REPORT_DATE	LAST_PURCHASE_DATE	LAST_PAYMENT_DATE	PAYMENT_FREQUENCY	BUSINESS_TYPE	CREDIT_TYPE	ACCOUNT_TYPE	RESPONSABILITY_TYPE	TOTAL_PAYMENTS	DELAYED_PAYMENTS	CURRENT_PAYMENT	WORST_DELAY	CURRENT_BALANCE	BALANCE_DUE	BALANCE_DUE_WORST_DELAY
0	4223	2022-04-01 00:32:36.689000+00:00	710278-27993a6e-2885-48d4-a753-9249e7022af1	2022-04-01	SIN OBSERVACION	PESOS MEXICANOS	1244.0	9900.0	1244.0	2022-02-28	2019-08-30	NaT	2022-01-31	2022-02-28	2019-08-30	2020-08-18	MENSUAL	TIENDA DEPARTAMENTAL	APARATOS/MUEBLES	PAGOS FIJOS	INDIVIDUAL (TITULAR)	12.0	7	13	13.0	1244.0	1244.0	1244.0
1	4223	2022-04-01 00:32:36.689000+00:00	710278-27993a6e-2885-48d4-a753-9249e7022af1	2022-04-01	SIN OBSERVACION	PESOS MEXICANOS	1722.0	9900.0	1722.0	2022-02-28	2020-04-21	NaT	2021-12-31	2022-02-28	2020-04-21	2020-08-18	PAGO MINIMO PARA CUENTAS REVOLVENTES	TIENDA DEPARTAMENTAL	LINEA DE CREDITO	REVOLVENTE	INDIVIDUAL (TITULAR)	0.0	3	13	13.0	1722.0	1722.0	1722.0
2	4223	2022-04-01 00:32:36.689000+00:00	710278-27993a6e-2885-48d4-a753-9249e7022af1	2022-04-01	SIN OBSERVACION	PESOS MEXICANOS	6346.0	0.0	482.0	2022-03-28	2021-12-17	NaT	NaT	2022-03-28	2021-12-17	2022-03-26	SEMANAL	BANCOS	PRESTAMO PERSONAL	PAGOS FIJOS	INDIVIDUAL (TITULAR)	16.0	0	V	0.0	943.0	0.0	0.0
3	3490	2022-02-15 02:30:22.086000+00:00	622857-6b4e9d95-7491-40c3-bccd-442de7f94c58	2022-02-15	SIN OBSERVACION	PESOS MEXICANOS	11600.0	0.0	232.0	2022-02-01	2021-03-02	NaT	2022-01-04	2022-02-01	2021-03-02	2022-01-25	SEMANAL	TIENDA DE ROPA	PRESTAMO PERSONAL	PAGOS FIJOS	INDIVIDUAL (TITULAR)	100.0	4	01	1.0	6185.0	116.0	116.0
4	6486	2022-06-25 01:57:14.868000+00:00	875073-46a5f149-19db-4193-b92a-b8f41fb3e896	2022-06-25	SIN OBSERVACION	PESOS MEXICANOS	2452.0	16800.0	2452.0	2021-09-30	2019-02-22	NaT	2020-08-31	2021-09-30	2019-04-16	2019-06-12	PAGO MINIMO PARA CUENTAS REVOLVENTES	TIENDA DEPARTAMENTAL	LINEA DE CREDITO	REVOLVENTE	INDIVIDUAL (TITULAR)	0.0	19	13	13.0	2452.0	2452.0	2452.0
5	6486	2022-06-25 01:57:14.868000+00:00	875073-46a5f149-19db-4193-b92a-b8f41fb3e896	2022-06-25	SIN OBSERVACION	PESOS MEXICANOS	0.0	0.0	0.0	2022-05-31	2021-11-30	NaT	NaT	2022-05-31	2022-05-16	2022-05-16	PAGO MINIMO PARA CUENTAS REVOLVENTES	SERVICIO DE TELEVISION DE PAGA	OTROS	SIN LIMITE PREESTABLECIDO	INDIVIDUAL (TITULAR)	0.0	0	V	0.0	0.0	0.0	0.0
6	4075	2022-03-13 01:46:41.593000+00:00	693356-d32e8877-2ba0-48c0-8366-08d733aa974e	2022-03-13	SIN OBSERVACION	PESOS MEXICANOS	7783.0	0.0	0.0	2004-09-23	2003-08-05	2004-04-27	NaT	2004-09-23	2003-08-05	2004-04-27	SEMANAL	MERCANCIA PARA HOGAR Y OFICINA	CREDITO AL CONSUMO	PAGOS FIJOS	INDIVIDUAL (TITULAR)	39.0	0	V	0.0	0.0	0.0	0.0

	count	mean	min	25%	50%	75%	max	std
customer_id	14454.0	7227.5	1.0	3614.25	7227.5	10840.75	14454.0	4172.654731
loan_id	14454.0	7227.5	1.0	3614.25	7227.5	10840.75	14454.0	4172.654731
ACC_CREATION_DATETIME	14454	2022-06-17 07:24:49.443337216	2020-10-14 18:22:10	2022-02-21 18:46:22.249999872	2022-07-19 20:29:43.500000	2022-11-13 07:37:39.249999872	2023-05-19 19:55:04	NaN
APPLICATION_DATETIME	14454	2022-11-28 03:42:40.896637696	2022-04-26 07:00:00	2022-09-15 13:00:00	2022-12-20 08:00:00	2023-02-04 08:00:00	2023-05-26 07:00:00	NaN
LOAN_ORIGINATION_DATETIME	14454	2022-12-28 06:04:09.504220416	2022-07-01 09:03:20	2022-10-27 21:15:58.249999872	2023-01-11 10:05:49.500000	2023-03-06 18:07:46.249999872	2023-05-29 12:18:28	NaN
max_days_late	14454.0	14.225889	-7.0	0.0	2.0	20.0	70.0	21.738445
target	14454.0	0.1868	0.0	0.0	0.0	0.0	1.0	0.389764
account_to_application_days	14454.0	163.489	0.0	0.0	103.0	271.75	901.0	181.110989
n_sf_apps	6806.0	1.653982	1.0	1.0	1.0	2.0	42.0	1.697131
first_app_date	6806	2022-06-16 04:31:39.970614272	2021-04-27 00:00:00	2022-02-27 00:00:00	2022-07-15 00:00:00	2022-10-21 00:00:00	2023-05-12 00:00:00	NaN
last_app_date	6806	2022-06-16 01:42:11.531002112	2021-04-25 00:00:00	2022-02-25 00:00:00	2022-07-16 00:00:00	2022-10-22 00:00:00	2023-05-12 00:00:00	NaN
n_bnpl_apps	8739.0	1.221765	1.0	1.0	1.0	1.0	18.0	0.831144
n_bnpl_approved_apps	8739.0	0.264904	0.0	0.0	0.0	0.0	15.0	0.602481
first_bnpl_app_date	8739	2022-08-13 09:30:46.852090880	2022-01-06 21:17:08.193000	2022-05-01 21:03:56.963500032	2022-08-18 13:36:14.271000064	2022-11-06 19:24:55.189499904	2023-05-20 17:15:47	NaN
last_bnpl_app_date	8739	2022-08-03 09:11:07.589265408	2022-01-06 21:17:08.193000	2022-04-20 05:33:33.585999872	2022-07-28 17:37:41.676999936	2022-11-06 01:50:47.641999872	2023-05-17 15:20:48	NaN
n_inquiries_l3m	9083.0	10.350435	0.0	0.0	0.0	14.0	170.0	19.694595
n_inquiries_l6m	9083.0	17.11483	0.0	0.0	8.0	26.0	213.0	23.229088

	count	mean	min	25%	50%	75%	max	std
customer_id	287356.0	6333.768075	1.0	2776.0	6187.0	9760.0	14416.0	3989.039684
INQUIRY_DATE	287267	2022-07-02 00:44:25.383771392	2021-04-29 00:00:00	2022-03-22 00:00:00	2022-07-18 00:00:00	2022-10-27 00:00:00	2023-05-17 00:00:00	NaN
MAX_CREDIT	287174.0	15280.383928	0.0	548.0	2300.0	7483.0	404040416.0	1313088.110101
CREDIT_LIMIT	278999.0	5525.841766	0.0	0.0	400.0	3000.0	1900000.0	27882.692051
PAYMENT_AMOUNT	287267.0	1670.948417	0.0	0.0	0.0	149.0	1800000.0	13010.321018
UPDATE_DATE	287267	2020-03-26 04:35:03.002433024	2001-04-25 00:00:00	2019-03-13 00:00:00	2021-11-22 00:00:00	2022-06-15 00:00:00	2023-05-07 00:00:00	NaN
LOAN_OPENING_DATE	287267	2018-11-22 15:24:26.870889984	1950-01-01 00:00:00	2017-03-21 00:00:00	2020-04-02 00:00:00	2021-10-14 00:00:00	2023-03-31 00:00:00	NaN
LOAN_CLOSING_DATE	192609	2018-10-24 21:45:09.910752	2000-08-28 00:00:00	2017-01-30 00:00:00	2020-01-27 00:00:00	2021-09-20 00:00:00	2023-03-31 00:00:00	NaN
WORST_DELAY_DATE	84658	2019-08-11 02:14:42.969123072	1998-10-10 00:00:00	2018-04-30 00:00:00	2021-05-31 00:00:00	2022-04-20 00:00:00	2023-04-12 00:00:00	NaN
REPORT_DATE	287267	2020-03-22 05:43:32.962853376	2001-04-25 00:00:00	2019-02-28 00:00:00	2021-11-10 00:00:00	2022-06-12 00:00:00	2023-05-07 00:00:00	NaN
LAST_PURCHASE_DATE	283214	2019-02-17 08:26:14.566228992	1950-01-01 00:00:00	2017-07-29 00:00:00	2020-08-03 00:00:00	2021-11-16 00:00:00	2023-04-07 00:00:00	NaN
LAST_PAYMENT_DATE	255679	2019-04-09 00:51:00.575174400	1950-01-01 00:00:00	2017-09-02 00:00:00	2020-09-06 00:00:00	2021-12-27 00:00:00	2023-05-07 00:00:00	NaN
TOTAL_PAYMENTS	268711.0	22.773876	0.0	1.0	3.0	16.0	1800.0	89.005749
WORST_DELAY	284146.0	4.27879	0.0	0.0	0.0	2.0	84.0	12.231546
CURRENT_BALANCE	287267.0	4578.482927	0.0	0.0	0.0	273.0	3469743.0	35543.251719
BALANCE_DUE	287267.0	2090.320639	0.0	0.0	0.0	0.0	1800000.0	14111.862995
BALANCE_DUE_WORST_DELAY	287267.0	1671.461922	0.0	0.0	0.0	159.0	1800000.0	11646.107604

	customer_id	loan_id	ACC_CREATION_DATETIME	APPLICATION_DATETIME	LOAN_ORIGINATION_DATETIME	max_days_late	account_to_application_days	n_sf_apps	first_app_date	last_app_date	n_bnpl_apps	first_bnpl_app_date	last_bnpl_app_date	n_inquiries_l3m	n_inquiries_l6m	INQUIRY_TIME	CDC_INQUIRY_ID	INQUIRY_DATE	PREVENTION_KEY	CURRENCY	MAX_CREDIT	CREDIT_LIMIT	UPDATE_DATE	LOAN_OPENING_DATE	LOAN_CLOSING_DATE	WORST_DELAY_DATE	REPORT_DATE	LAST_PURCHASE_DATE	LAST_PAYMENT_DATE	PAYMENT_FREQUENCY	BUSINESS_TYPE	CREDIT_TYPE	ACCOUNT_TYPE	RESPONSABILITY_TYPE	TOTAL_PAYMENTS	CURRENT_PAYMENT	WORST_DELAY	TOTAL_REPORTED_PAYMENTS
0	1223	1	2021-08-23 13:57:56	2022-04-26 07:00:00	2022-10-20 18:26:35	5	245	1.0	2022-04-25	2022-04-25	1.0	2022-04-26 02:15:17.742	2022-04-26 02:15:17.742	91.0	91.0	2022-04-25 21:58:16.422000+00:00	221105-252f8dc6-04b8-41dc-92c3-bd9902478b8a	2022-04-25	SIN OBSERVACION	PESOS MEXICANOS	3000.0	0.0	2017-04-19	2013-03-27	2017-02-21	NaT	2017-02-28	2017-02-21	2013-03-27	MENSUAL	SERVICIOS	LINEA DE CREDITO	SIN LIMITE PREESTABLECIDO	INDIVIDUAL (TITULAR)	NaN	V	NaN	None
1	1223	1	2021-08-23 13:57:56	2022-04-26 07:00:00	2022-10-20 18:26:35	5	245	1.0	2022-04-25	2022-04-25	1.0	2022-04-26 02:15:17.742	2022-04-26 02:15:17.742	91.0	91.0	2022-04-25 21:58:16.422000+00:00	221105-252f8dc6-04b8-41dc-92c3-bd9902478b8a	2022-04-25	SIN OBSERVACION	PESOS MEXICANOS	500.0	2200.0	2021-06-21	2021-04-03	2021-04-18	NaT	2021-06-21	2021-04-03	2021-04-18	MENSUAL	COMPANIA DE PRESTAMO PERSONAL	PRESTAMO PERSONAL	PAGOS FIJOS	INDIVIDUAL (TITULAR)	1.0	V	0.0	0
2	1223	1	2021-08-23 13:57:56	2022-04-26 07:00:00	2022-10-20 18:26:35	5	245	1.0	2022-04-25	2022-04-25	1.0	2022-04-26 02:15:17.742	2022-04-26 02:15:17.742	91.0	91.0	2022-04-25 21:58:16.422000+00:00	221105-252f8dc6-04b8-41dc-92c3-bd9902478b8a	2022-04-25	SIN OBSERVACION	PESOS MEXICANOS	1100.0	2200.0	2021-06-21	2021-04-18	2021-04-22	NaT	2021-06-21	2021-04-18	2021-04-22	MENSUAL	COMPANIA DE PRESTAMO PERSONAL	PRESTAMO PERSONAL	PAGOS FIJOS	INDIVIDUAL (TITULAR)	1.0	V	0.0	0
3	1223	1	2021-08-23 13:57:56	2022-04-26 07:00:00	2022-10-20 18:26:35	5	245	1.0	2022-04-25	2022-04-25	1.0	2022-04-26 02:15:17.742	2022-04-26 02:15:17.742	91.0	91.0	2022-04-25 21:58:16.422000+00:00	221105-252f8dc6-04b8-41dc-92c3-bd9902478b8a	2022-04-25	SIN OBSERVACION	PESOS MEXICANOS	1700.0	2500.0	2021-07-13	2021-04-22	2021-05-03	NaT	2021-07-13	2021-04-22	2021-05-03	MENSUAL	COMPANIA DE PRESTAMO PERSONAL	PRESTAMO PERSONAL	PAGOS FIJOS	INDIVIDUAL (TITULAR)	1.0	V	0.0	0
4	1223	1	2021-08-23 13:57:56	2022-04-26 07:00:00	2022-10-20 18:26:35	5	245	1.0	2022-04-25	2022-04-25	1.0	2022-04-26 02:15:17.742	2022-04-26 02:15:17.742	91.0	91.0	2022-04-25 21:58:16.422000+00:00	221105-252f8dc6-04b8-41dc-92c3-bd9902478b8a	2022-04-25	SIN OBSERVACION	PESOS MEXICANOS	1800.0	2500.0	2021-07-13	2021-05-03	2021-05-26	NaT	2021-07-13	2021-05-03	2021-05-26	MENSUAL	COMPANIA DE PRESTAMO PERSONAL	PRESTAMO PERSONAL	PAGOS FIJOS	INDIVIDUAL (TITULAR)	1.0	V	0.0	0

Table of Contents¶

Business Objective¶

Basic statistics¶

Merging Datasets¶

Exploratory Data Analysis¶

Univariate Analysis¶

Univariate Analysis of Categorical Variables¶

bivariate-multivariate-analysis¶

Data Preparation and Feature Engineering¶

feature engineering¶

Kmeans Clustering¶

RFM Analysis¶

Data Pipeline Tranformation¶

Model Development¶

Trainig and Test Oversampling SMOTE¶

Model Selection¶

Model Evaluation¶

Model Interpretation¶

Feature Importance¶

Results and Recommendations¶

	max_credit	time_to_loan_days	gen_credit_type	credit_limit	update_date	n_inquiries_l3m	prevention_key	n_bnpl_apps	outstanding_balance	inquiry_date	last_app_date	n_sf_apps	loan_origination_datetime	last_payment_days	worst_delay_date	account_type	first_app_date	payment_frequency	account_to_application_days	first_bnpl_app_date	n_inquiries_l6m	acc_creation_datetime	loan_opening_date	total_payments	report_date	responsability_type	max_days_late	last_bnpl_app_date	loan_closing_date	currency	last_purchase_date	loan_life_days	customer_id	gen_business_type	application_datetime	last_payment_date
0	3000.0	178	Miscellaneous	0.0	2017-04-19	91.0	SIN OBSERVACION	1.0	0.0	2022-04-25	2022-04-25	1.0	2022-10-20 18:26:35	1427.0	NaT	SIN LIMITE PREESTABLECIDO	2022-04-25	MENSUAL	245	2022-04-26 02:15:17.742	91.0	2021-08-23 13:57:56	2013-03-27	NaN	2017-02-28	INDIVIDUAL (TITULAR)	5	2022-04-26 02:15:17.742	2017-02-21	PESOS MEXICANOS	2017-02-21	1427.0	1223	General Services	2022-04-26 07:00:00	2013-03-27
1	500.0	178	Personal Loans	2200.0	2021-06-21	91.0	SIN OBSERVACION	1.0	2200.0	2022-04-25	2022-04-25	1.0	2022-10-20 18:26:35	15.0	NaT	PAGOS FIJOS	2022-04-25	MENSUAL	245	2022-04-26 02:15:17.742	91.0	2021-08-23 13:57:56	2021-04-03	1.0	2021-06-21	INDIVIDUAL (TITULAR)	5	2022-04-26 02:15:17.742	2021-04-18	PESOS MEXICANOS	2021-04-03	15.0	1223	Financial Services	2022-04-26 07:00:00	2021-04-18
2	1100.0	178	Personal Loans	2200.0	2021-06-21	91.0	SIN OBSERVACION	1.0	2200.0	2022-04-25	2022-04-25	1.0	2022-10-20 18:26:35	4.0	NaT	PAGOS FIJOS	2022-04-25	MENSUAL	245	2022-04-26 02:15:17.742	91.0	2021-08-23 13:57:56	2021-04-18	1.0	2021-06-21	INDIVIDUAL (TITULAR)	5	2022-04-26 02:15:17.742	2021-04-22	PESOS MEXICANOS	2021-04-18	4.0	1223	Financial Services	2022-04-26 07:00:00	2021-04-22
3	1700.0	178	Personal Loans	2500.0	2021-07-13	91.0	SIN OBSERVACION	1.0	2500.0	2022-04-25	2022-04-25	1.0	2022-10-20 18:26:35	11.0	NaT	PAGOS FIJOS	2022-04-25	MENSUAL	245	2022-04-26 02:15:17.742	91.0	2021-08-23 13:57:56	2021-04-22	1.0	2021-07-13	INDIVIDUAL (TITULAR)	5	2022-04-26 02:15:17.742	2021-05-03	PESOS MEXICANOS	2021-04-22	11.0	1223	Financial Services	2022-04-26 07:00:00	2021-05-03
4	1800.0	178	Personal Loans	2500.0	2021-07-13	91.0	SIN OBSERVACION	1.0	2500.0	2022-04-25	2022-04-25	1.0	2022-10-20 18:26:35	23.0	NaT	PAGOS FIJOS	2022-04-25	MENSUAL	245	2022-04-26 02:15:17.742	91.0	2021-08-23 13:57:56	2021-05-03	1.0	2021-07-13	INDIVIDUAL (TITULAR)	5	2022-04-26 02:15:17.742	2021-05-26	PESOS MEXICANOS	2021-05-03	23.0	1223	Financial Services	2022-04-26 07:00:00	2021-05-26

	customer_id	delayed_payments_mean	delayed_payments_sum	delayed_payments_count	max_credit_mean	max_credit_sum	max_credit_count	time_to_loan_days_mean	time_to_loan_days_sum	time_to_loan_days_count	credit_limit_mean	credit_limit_sum	credit_limit_count	update_date_min	update_date_max	n_inquiries_l3m_mean	n_inquiries_l3m_sum	n_inquiries_l3m_count	prevention_key_nunique	current_balance_mean	current_balance_sum	current_balance_count	n_bnpl_apps_mean	n_bnpl_apps_sum	n_bnpl_apps_count	outstanding_balance_mean	outstanding_balance_sum	outstanding_balance_count	inquiry_date_min	inquiry_date_max	last_app_date_min	last_app_date_max	n_sf_apps_mean	n_sf_apps_sum	n_sf_apps_count	loan_origination_datetime_min	loan_origination_datetime_max	last_payment_days_mean	last_payment_days_sum	last_payment_days_count	payment_to_due_ratio_mean	payment_to_due_ratio_sum	payment_to_due_ratio_count	worst_delay_date_min	worst_delay_date_max	first_app_date_min	first_app_date_max	n_bnpl_approved_apps_mean	n_bnpl_approved_apps_sum	n_bnpl_approved_apps_count	payment_amount_mean	payment_amount_sum	payment_amount_count	account_to_application_days_mean	account_to_application_days_sum	account_to_application_days_count	first_bnpl_app_date_min	first_bnpl_app_date_max	n_inquiries_l6m_mean	n_inquiries_l6m_sum	n_inquiries_l6m_count	acc_creation_datetime_min	acc_creation_datetime_max	loan_opening_date_min	loan_opening_date_max	total_payments_mean	total_payments_sum	total_payments_count	report_date_min	report_date_max	max_days_late_mean	max_days_late_sum	max_days_late_count	target_mean	target_sum	target_count	last_bnpl_app_date_min	last_bnpl_app_date_max	loan_closing_date_min	loan_closing_date_max	last_purchase_date_min	last_purchase_date_max	loan_life_days_mean	loan_life_days_sum	loan_life_days_count	application_datetime_min	application_datetime_max	last_payment_date_min	last_payment_date_max	account_type_CREDITO DE HABILITACION O AVIO_mean	account_type_CREDITO DE HABILITACION O AVIO_sum	account_type_CREDITO DE HABILITACION O AVIO_count	account_type_HIPOTECA_mean	account_type_HIPOTECA_sum	account_type_HIPOTECA_count	account_type_PAGOS FIJOS_mean	account_type_PAGOS FIJOS_sum	account_type_PAGOS FIJOS_count	account_type_POR DETERMINAR_mean	account_type_POR DETERMINAR_sum	account_type_POR DETERMINAR_count	account_type_REVOLVENTE_mean	account_type_REVOLVENTE_sum	account_type_REVOLVENTE_count	account_type_SIN LIMITE PREESTABLECIDO_mean	account_type_SIN LIMITE PREESTABLECIDO_sum	account_type_SIN LIMITE PREESTABLECIDO_count	currency_DOLARES_mean	currency_DOLARES_sum	currency_DOLARES_count	currency_PESOS MEXICANOS_mean	currency_PESOS MEXICANOS_sum	currency_PESOS MEXICANOS_count	currency_UNIDADES DE INVERSION_mean	currency_UNIDADES DE INVERSION_sum	currency_UNIDADES DE INVERSION_count	responsability_type_AVAL_mean	responsability_type_AVAL_sum	responsability_type_AVAL_count	responsability_type_INDIVIDUAL (TITULAR)_mean	responsability_type_INDIVIDUAL (TITULAR)_sum	responsability_type_INDIVIDUAL (TITULAR)_count	responsability_type_MANCOMUNADO_mean	responsability_type_MANCOMUNADO_sum	responsability_type_MANCOMUNADO_count	responsability_type_OBLIGATORIO SOLIDARIO_mean	responsability_type_OBLIGATORIO SOLIDARIO_sum	responsability_type_OBLIGATORIO SOLIDARIO_count	responsability_type_TITULAR CON AVAL_mean	responsability_type_TITULAR CON AVAL_sum	responsability_type_TITULAR CON AVAL_count	gen_credit_type_mean	gen_credit_type_sum	gen_credit_type_count	gen_business_type_mean	gen_business_type_sum	gen_business_type_count	payment_frequency_mean	payment_frequency_sum	payment_frequency_count	prevention_key_risk_score_mean	prevention_key_risk_score_sum	prevention_key_risk_score_count	customer_lifetime
0	1	0.125000	1	8	21257.000000	170056.0	8	4.0	32	8	69080.000000	552640.0	8	2018-01-24	2022-06-11	NaN	0.0	0	2	2246.000000	17968.0	8	2.0	16.0	8	66834.000000	534672.0	8	2022-06-16	2022-06-16	2022-06-16	2022-06-16	1.0	8.0	8	2023-02-16 11:44:50	2023-02-16 11:44:50	76.250000	610.0	8	4.272500	34.18	8	NaT	NaT	2022-06-16	2022-06-16	2.0	16.0	8	202.625000	1621.0	8	851.0	6808	8	2022-10-27 19:15:56.544	2022-10-27 19:15:56.544	NaN	0.0	0	2020-10-14 18:22:10	2020-10-14 18:22:10	2003-08-21	2017-04-21	1.000000	1.0	1	2017-11-30	2022-05-31	0.0	0	8	0.0	0	8	2022-07-21 17:38:17.667	2022-07-21 17:38:17.667	2017-12-19	2018-04-05	2017-11-30	2022-05-31	1341.500000	2683.0	2	2023-02-13 08:00:00	2023-02-13 08:00:00	2017-11-28	2022-05-27	0.0	0	8	0.00000	0	8	0.000000	0	8	0.000000	0	8	0.750000	6	8	0.250000	2	8	0.0	0	8	1.0	8	8	0.0	0	8	0.0	0	8	1.000000	8	8	0.000000	0	8	0.000000	0	8	0.000000	0	8	0.178660	1.429281	8	0.173792	1.390340	8	0.162429	1.299428	8	1.250000	10.0	8	851
1	2	0.000000	0	10	15227.500000	152275.0	10	2.0	20	10	37566.666667	338100.0	9	2020-04-09	2022-04-25	0.0	0.0	10	4	2577.900000	25779.0	10	1.0	10.0	10	34702.333333	312321.0	9	2022-05-02	2022-05-02	2022-11-09	2022-11-09	8.0	80.0	10	2023-01-27 16:17:49	2023-01-27 16:17:49	242.500000	2425.0	10	NaN	NaN	10	NaT	NaT	2022-05-02	2022-05-02	1.0	10.0	10	168.900000	1689.0	10	828.0	8280	10	2022-07-01 16:50:01.030	2022-07-01 16:50:01.030	0.0	0.0	10	2020-10-19 12:11:43	2020-10-19 12:11:43	2009-08-15	2021-07-26	18.500000	37.0	2	2020-03-31	2022-04-19	-1.0	-10	10	0.0	0	10	2022-07-01 16:50:01.030	2022-07-01 16:50:01.030	2020-03-11	2022-01-21	2016-01-05	2022-04-14	1256.333333	7538.0	6	2023-01-26 08:00:00	2023-01-26 08:00:00	2019-11-26	2022-04-16	0.0	0	10	0.00000	0	10	0.100000	1	10	0.000000	0	10	0.900000	9	10	0.000000	0	10	0.0	0	10	1.0	10	10	0.0	0	10	0.0	0	10	1.000000	10	10	0.000000	0	10	0.000000	0	10	0.000000	0	10	0.183405	1.834049	10	0.165376	1.653760	10	0.162429	1.624285	10	2.000000	20.0	10	828
2	3	2.656250	85	32	14743.343750	471787.0	32	11.0	352	32	16751.250000	536040.0	32	2014-06-16	2022-10-12	0.0	0.0	32	4	14773.718750	472759.0	32	1.0	32.0	32	3392.906250	108573.0	32	2022-10-19	2022-10-19	NaT	NaT	NaN	0.0	0	2023-02-02 18:21:12	2023-02-02 18:21:12	91.269231	2373.0	26	4.142500	132.56	32	2021-12-31	2022-07-31	NaT	NaT	0.0	0.0	32	379.468750	12143.0	32	818.0	26176	32	2022-10-19 03:51:19.918	2022-10-19 03:51:19.918	31.0	992.0	32	2020-10-26 11:03:57	2020-10-26 11:03:57	2014-05-03	2021-12-06	12.812500	410.0	32	2014-06-16	2022-09-30	63.0	2016	32	1.0	32	32	2022-10-19 03:51:19.918	2022-10-19 03:51:19.918	2014-06-16	2022-08-31	2014-05-03	2021-12-06	67.090909	1476.0	22	2023-01-23 08:00:00	2023-01-23 08:00:00	2014-06-16	2022-09-07	0.0	0	32	0.03125	1	32	0.906250	29	32	0.062500	2	32	0.000000	0	32	0.000000	0	32	0.0	0	32	1.0	32	32	0.0	0	32	0.0	0	32	1.000000	32	32	0.000000	0	32	0.000000	0	32	0.000000	0	32	0.165853	5.307290	32	0.167984	5.375483	32	0.163515	5.232486	32	1.500000	48.0	32	818
3	4	0.153846	2	13	3911.923077	50855.0	13	23.0	299	13	2858.076923	37155.0	13	2020-10-07	2022-01-11	0.0	0.0	13	2	3447.923077	44823.0	13	7.0	91.0	13	2155.846154	28026.0	13	2022-01-14	2022-01-14	2022-01-21	2022-01-21	6.0	78.0	13	2022-10-06 13:07:47	2022-10-06 13:07:47	49.181818	541.0	11	4.870769	63.32	13	2021-11-30	2021-11-30	2022-01-14	2022-01-14	0.0	0.0	13	421.307692	5477.0	13	687.0	8931	13	2022-03-14 01:34:34.104	2022-03-14 01:34:34.104	0.0	0.0	13	2020-10-26 12:47:09	2020-10-26 12:47:09	2020-06-30	2021-12-15	5.900000	59.0	10	2020-09-30	2022-01-03	28.0	364	13	0.0	0	13	2022-02-19 22:16:14.775	2022-02-19 22:16:14.775	2020-09-01	2021-12-15	2020-06-30	2021-12-26	26.333333	158.0	6	2022-09-14 07:00:00	2022-09-14 07:00:00	2020-09-01	2022-01-03	0.0	0	13	0.00000	0	13	0.769231	10	13	0.000000	0	13	0.230769	3	13	0.000000	0	13	0.0	0	13	1.0	13	13	0.0	0	13	0.0	0	13	1.000000	13	13	0.000000	0	13	0.000000	0	13	0.000000	0	13	0.169715	2.206294	13	0.169220	2.199859	13	0.164236	2.135074	13	1.076923	14.0	13	687
4	5	6.204082	304	49	27515.775510	1348273.0	49	28.0	1372	49	18409.478261	846836.0	46	2007-08-01	2021-10-01	0.0	0.0	49	9	12025.612245	589255.0	49	NaN	0.0	0	10717.413043	493001.0	46	2021-10-04	2021-10-04	2021-10-04	2021-10-04	1.0	49.0	49	2022-10-18 13:12:47	2022-10-18 13:12:47	267.133333	12021.0	45	2.582857	126.56	49	2007-07-29	2021-08-31	2021-10-04	2021-10-04	NaN	0.0	0	9184.857143	450058.0	49	694.0	34006	49	NaT	NaT	0.0	0.0	49	2020-10-26 20:13:57	2020-10-26 20:13:57	2002-06-18	2021-05-14	48.621622	1799.0	37	2007-08-01	2021-09-30	6.0	294	49	0.0	0	49	NaT	NaT	2007-08-01	2021-08-17	2007-03-25	2021-09-04	528.000000	14256.0	27	2022-09-21 07:00:00	2022-09-21 07:00:00	2007-08-01	2021-09-24	0.0	0	49	0.00000	0	49	0.408163	20	49	0.000000	0	49	0.367347	18	49	0.224490	11	49	0.0	0	49	1.0	49	49	0.0	0	49	0.0	0	49	1.000000	49	49	0.000000	0	49	0.000000	0	49	0.000000	0	49	0.172096	8.432726	49	0.165770	8.122741	49	0.166272	8.147327	49	3.142857	154.0	49	694
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
9244	14391	2.538462	66	26	18744.423077	487355.0	26	7.0	182	26	4722.269231	122779.0	26	2019-10-31	2023-04-06	52.0	1352.0	26	3	541.576923	14081.0	26	1.0	26.0	26	5141.538462	133680.0	26	2023-04-21	2023-04-21	2023-04-20	2023-04-20	1.0	26.0	26	2023-04-27 15:08:48	2023-04-27 15:08:48	106.333333	2552.0	24	0.192308	5.00	26	2020-02-29	2023-02-28	2023-04-20	2023-04-20	0.0	0.0	26	541.576923	14081.0	26	0.0	0	26	2023-04-21 01:37:54.043	2023-04-21 01:37:54.043	52.0	1352.0	26	2023-04-20 20:01:02	2023-04-20 20:01:02	2019-05-31	2022-08-26	15.000000	345.0	23	2019-10-31	2023-03-31	21.0	546	26	0.0	0	26	2023-04-21 01:37:54.043	2023-04-21 01:37:54.043	2019-10-14	2023-03-30	2019-06-06	2023-03-01	212.571429	4464.0	21	2023-04-21 07:00:00	2023-04-21 07:00:00	2019-10-13	2023-03-01	0.0	0	26	0.00000	0	26	0.730769	19	26	0.038462	1	26	0.076923	2	26	0.153846	4	26	0.0	0	26	1.0	26	26	0.0	0	26	0.0	0	26	0.807692	21	26	0.038462	1	26	0.153846	4	26	0.000000	0	26	0.170023	4.420607	26	0.170693	4.438015	26	0.168942	4.392479	26	2.461538	64.0	26	0
9245	14404	9.115385	237	26	4369.269231	113601.0	26	3.0	78	26	3129.653846	81371.0	26	2005-12-18	2023-04-19	28.0	728.0	26	5	2609.807692	67855.0	26	1.0	26.0	26	2935.692308	76328.0	26	2023-04-23	2023-04-23	NaT	NaT	NaN	0.0	0	2023-04-27 17:24:48	2023-04-27 17:24:48	71.208333	1709.0	24	1.433077	37.26	26	2005-11-01	2023-03-31	NaT	NaT	0.0	0.0	26	2249.076923	58476.0	26	2.0	52	26	2023-04-23 06:05:02.883	2023-04-23 06:05:02.883	28.0	728.0	26	2023-04-22 20:37:42	2023-04-22 20:37:42	2005-05-31	2023-02-28	16.416667	394.0	24	2005-12-18	2023-04-19	-1.0	-26	26	0.0	0	26	2023-04-23 06:05:02.883	2023-04-23 06:05:02.883	2005-12-18	2023-03-06	2005-05-31	2023-03-11	128.937500	2063.0	16	2023-04-25 07:00:00	2023-04-25 07:00:00	2005-12-18	2023-04-10	0.0	0	26	0.00000	0	26	0.769231	20	26	0.000000	0	26	0.115385	3	26	0.115385	3	26	0.0	0	26	1.0	26	26	0.0	0	26	0.0	0	26	1.000000	26	26	0.000000	0	26	0.000000	0	26	0.000000	0	26	0.167796	4.362688	26	0.167519	4.355506	26	0.167344	4.350938	26	2.923077	76.0	26	2
9246	14408	35.250000	141	4	29629.000000	118516.0	4	4.0	16	4	0.000000	0.0	4	2021-02-28	2023-04-09	10.0	40.0	4	2	7764.250000	31057.0	4	1.0	4.0	4	7764.250000	31057.0	4	2023-04-23	2023-04-23	2023-04-23	2023-04-23	1.0	4.0	4	2023-04-27 09:32:32	2023-04-27 09:32:32	83.000000	332.0	4	0.500000	2.00	4	2020-12-20	2021-03-13	2023-04-23	2023-04-23	0.0	0.0	4	7764.250000	31057.0	4	0.0	0	4	2023-04-23 18:06:58.837	2023-04-23 18:06:58.837	10.0	40.0	4	2023-04-23 12:53:02	2023-04-23 12:53:02	2019-03-10	2021-02-22	63.500000	254.0	4	2021-02-28	2023-04-09	23.0	92	4	0.0	0	4	2023-04-23 18:06:58.837	2023-04-23 18:06:58.837	2021-02-17	2021-06-30	2019-03-10	2021-02-22	113.500000	227.0	2	2023-04-24 07:00:00	2023-04-24 07:00:00	2019-04-27	2021-06-30	0.0	0	4	0.00000	0	4	1.000000	4	4	0.000000	0	4	0.000000	0	4	0.000000	0	4	0.0	0	4	1.0	4	4	0.0	0	4	0.0	0	4	1.000000	4	4	0.000000	0	4	0.000000	0	4	0.000000	0	4	0.168101	0.672403	4	0.162282	0.649126	4	0.170136	0.680544	4	5.500000	22.0	4	0
9247	14410	2.262821	353	156	2796.339744	436229.0	156	3.0	468	156	123.384615	19248.0	156	2011-07-27	2023-04-19	33.0	5148.0	156	5	183.089744	28562.0	156	1.0	156.0	156	209.846154	32736.0	156	2023-04-23	2023-04-23	2023-04-23	2023-04-23	1.0	156.0	156	2023-04-26 21:25:09	2023-04-26 21:25:09	89.793333	13469.0	150	0.160256	25.00	156	2011-07-20	2023-04-12	2023-04-23	2023-04-23	0.0	0.0	156	168.448718	26278.0	156	0.0	0	156	2023-04-23 20:54:13.506	2023-04-23 20:54:13.506	33.0	5148.0	156	2023-04-23 15:49:12	2023-04-23 15:49:12	2011-03-25	2023-02-19	4.993506	769.0	154	2011-07-27	2023-04-19	0.0	0	156	0.0	0	156	2023-04-23 20:54:13.506	2023-04-23 20:54:13.506	2011-07-22	2023-02-22	2011-03-25	2023-03-24	85.902098	12284.0	143	2023-04-24 07:00:00	2023-04-24 07:00:00	2011-07-21	2023-03-22	0.0	0	156	0.00000	0	156	0.237179	37	156	0.730769	114	156	0.025641	4	156	0.006410	1	156	0.0	0	156	1.0	156	156	0.0	0	156	0.0	0	156	0.974359	152	156	0.000000	0	156	0.012821	2	156	0.012821	2	156	0.166609	25.991020	156	0.158096	24.663003	156	0.169450	26.434217	156	1.256410	196.0	156	0
9248	14416	1.821429	51	28	20347.642857	569734.0	28	8.0	224	28	21060.392857	589691.0	28	2009-04-03	2023-04-02	11.0	308.0	28	3	5381.107143	150671.0	28	1.0	28.0	28	15679.285714	439020.0	28	2023-04-24	2023-04-24	2023-04-24	2023-04-24	1.0	28.0	28	2023-05-05 21:09:11	2023-05-05 21:09:11	223.400000	5585.0	25	0.107143	3.00	28	2008-07-30	2023-03-31	2023-04-24	2023-04-24	0.0	0.0	28	5381.107143	150671.0	28	3.0	84	28	2023-04-24 23:36:58.288	2023-04-24 23:36:58.288	11.0	308.0	28	2023-04-24 18:36:17	2023-04-24 18:36:17	2005-11-04	2023-03-17	11.925926	322.0	27	2009-04-03	2023-04-02	0.0	0	28	0.0	0	28	2023-04-24 23:36:58.288	2023-04-24 23:36:58.288	2006-10-31	2023-02-22	2005-11-04	2021-12-30	587.960000	14699.0	25	2023-04-28 07:00:00	2023-04-28 07:00:00	2006-10-31	2023-01-12	0.0	0	28	0.00000	0	28	0.678571	19	28	0.107143	3	28	0.142857	4	28	0.071429	2	28	0.0	0	28	1.0	28	28	0.0	0	28	0.0	0	28	1.000000	28	28	0.000000	0	28	0.000000	0	28	0.000000	0	28	0.157965	4.423030	28	0.157886	4.420809	28	0.165593	4.636598	28	1.357143	38.0	28	3

	customer_id	target	customer_lifetime	n_sf_apps	prevention_key_nunique	total_credits	loan_life_days_mean	max_credit	total_payments_sum	n_inquiries_l3m_mean	payment_amount	payment_to_due_ratio	time_to_loan_days_mean	credit_limit	n_bnpl_approved_apps_mean	outstanding_balance	n_inquiries_l6m_mean	n_bnpl_apps	max_days_late	last_payment_days	delayed_payments	current_balance	account_type_pagos fijos_sum	account_type_revolvente_sum	account_type_sin limite preestablecido_sum	currency_pesos mexicanos_sum	responsability_type_individual (titular)_sum	prevention_key_risk_score	gen_credit_type	gen_business_type	payment_frequency	Lifetime_Cluster
3535	4290	1.0	190	28.0	3	28	44.923077	47222.0	102.0	0.0	15731.0	2.977500	2.0	17106.0	0.0	601.571429	0.0	56.0	1736	964.0	177	30148.0	21	3	4	28	28	1.321429	0.168919	0.166691	0.163791	Medium_Tenure
8470	11758	0.0	8	13.0	4	13	217.750000	71356.0	149.0	16.0	3377.0	10.236154	13.0	24469.0	0.0	3162.000000	16.0	13.0	13	1008.0	49	19089.0	7	0	6	13	13	3.076923	0.168947	0.175107	0.166437	Long_Tenure
3862	4692	0.0	251	4.0	2	4	NaN	64351.0	906.0	0.0	86493.0	9.342500	8.0	51500.0	1.0	11773.000000	0.0	8.0	24	74.0	25	98572.0	1	3	0	4	4	3.000000	0.173848	0.166270	0.166254	Medium_Tenure
9019	13611	0.0	0	0.0	7	60	143.975610	225496.0	313.0	70.0	244416.0	0.696000	5.0	166976.0	NaN	4564.859649	70.0	0.0	0	3992.0	218	253137.0	49	9	2	60	60	2.016667	0.168732	0.168494	0.166905	Long_Tenure
5797	7404	0.0	204	0.0	4	21	145.750000	192926.0	1162.0	0.0	21165.0	0.341905	3.0	147853.0	0.0	5801.047619	0.0	21.0	0	2362.0	52	26031.0	17	4	0	21	21	2.047619	0.167250	0.167660	0.167200	Medium_Tenure

	Recency	Frequency	Monetary
count	9145.000000	9249.000000	9172.000000
mean	463.319956	30.839766	6852.744311
std	424.451785	30.582474	14799.353291
min	0.000000	1.000000	0.000000
25%	249.000000	10.000000	1223.190000
50%	374.000000	21.000000	2614.046296
75%	542.000000	41.000000	5960.183279
max	6474.000000	269.000000	481599.500000

	customer_id	Recent_Payment_Date	Recency	Frequency	Monetary
7671	10605	2022-10-17	202.0	28	1878.642857
73	90	2022-06-10	331.0	91	432.857143
8602	12060	2022-12-19	139.0	19	2620.000000
8324	11564	2022-11-18	170.0	9	226.555556
8871	13084	2023-01-09	118.0	139	3130.467626

	RFM_Level	Count
0	Average	6168
1	At Risk	2222
2	Needs Attention	510
3	Loyal Customer	222
4	Champion	127

	customer_id	target	customer_lifetime	n_sf_apps	prevention_key_nunique	total_credits	loan_life_days_mean	max_credit	total_payments_sum	n_inquiries_l3m_mean	payment_amount	payment_to_due_ratio	time_to_loan_days_mean	credit_limit	n_bnpl_approved_apps_mean	outstanding_balance	n_inquiries_l6m_mean	n_bnpl_apps	max_days_late	last_payment_days	delayed_payments	current_balance	account_type_hipoteca_sum	account_type_pagos fijos_sum	account_type_por determinar_sum	account_type_revolvente_sum	account_type_sin limite preestablecido_sum	currency_pesos mexicanos_sum	responsability_type_individual (titular)_sum	responsability_type_mancomunado_sum	responsability_type_obligatorio solidario_sum	responsability_type_titular con aval_sum	prevention_key_risk_score	gen_credit_type	gen_business_type	payment_frequency	Lifetime_Cluster	RFM_Score
1953	2269	0.0	538	242.0	5	121	121.812500	514592.0	754.0	0.0	196207.0	0.791488	6.0	286582.0	0.0	2498.965217	0.0	121.0	3267	14090.0	285	246840.0	0	73	37	7	4	121	82	1	0	38	1.685950	0.166163	0.164210	0.168463	Short_Tenure	142
2918	3525	0.0	243	19.0	2	19	40.166667	98732.0	241.0	0.0	32347.0	1.493158	4.0	23265.0	0.0	2709.368421	0.0	19.0	228	2043.0	178	33943.0	0	6	10	1	2	19	19	0	0	0	2.263158	0.166774	0.161613	0.168102	Medium_Tenure	122
6528	8768	1.0	143	1.0	1	1	252.000000	5213.0	16.0	0.0	3558.0	1.000000	30.0	0.0	0.0	3558.000000	8.0	1.0	42	252.0	9	3558.0	0	1	0	0	0	1	1	0	0	0	9.000000	0.166172	0.169220	0.170136	Long_Tenure	213
2842	3429	0.0	318	114.0	4	57	164.217391	765839.0	1426.0	0.0	139123.0	5.715088	23.0	474741.0	1.0	1334.272727	0.0	57.0	1596	5605.0	277	503648.0	2	53	0	1	1	57	55	0	2	0	2.561404	0.167614	0.166786	0.167638	Medium_Tenure	244
4817	5967	1.0	64	0.0	4	29	185.454545	103264.0	390.0	26.0	21002.0	0.922069	107.0	30888.0	0.0	1632.827586	26.0	29.0	1798	3987.0	34	23588.0	0	25	0	1	3	29	29	0	0	0	1.793103	0.167236	0.171364	0.167094	Long_Tenure	331

	Classifier	AUC_train	AUC_CV_train	AUC_validation	AUC_test
0	RF	0.982555	0.999646	0.995155	0.940994
1	AdaBoost	0.987539	0.999339	0.998176	0.979315
2	GBoosting	1.000000	1.000000	0.999153	0.994111
3	XGBoost	1.000000	1.000000	0.999783	0.998671
4	Logistic_r	0.963449	0.990104	0.989697	0.951683
5	D_tree	0.990281	0.999219	0.994379	0.975179